Michael I. Jordan: Machine Learning, Recommender Systems, and the Future of AI

Lex Fridman (02:18.480)

me an algorithmic marvel.

Michael I. Jordan (02:21.200)

Great props for the Cash App engineers for solving a hard problem that in the end provides

Michael I. Jordan (02:26.400)

an easy interface that takes a step up to the next layer of abstraction over the stock

Michael I. Jordan (02:31.140)

market, making trading more accessible for new investors and diversification much easier.

Lex Fridman (02:38.440)

So once again, if you get Cash App from the App Store or Google Play and use the code

Michael I. Jordan (02:42.760)

LEX PODCAST, you'll get $10 and Cash App will also donate $10 to First, one of my favorite

Michael I. Jordan (02:49.280)

organizations that is helping to advance robotics and STEM education for young people around

Michael I. Jordan (02:55.120)

the world.

Lex Fridman (02:57.120)

And now, here's my conversation with Michael I. Jordan.

Michael I. Jordan (03:02.760)

Given that you're one of the greats in the field of AI, machine learning, computer science,

Lex Fridman (03:06.480)

and so on, you're trivially called the Michael Jordan of machine learning, although as you

Michael I. Jordan (03:14.000)

know, you were born first, so technically MJ is the Michael I. Jordan of basketball.

Lex Fridman (03:19.680)

But anyway, my favorite is Yann LeCun calling you the Miles Davis of machine learning because

Michael I. Jordan (03:25.520)

as he says, you reinvent yourself periodically and sometimes leave fans scratching their

Lex Fridman (03:30.600)

heads after you change direction.

Lex Fridman (03:32.460)

So can you put at first your historian hat on and give a history of computer science

Lex Fridman (03:38.840)

and AI as you saw it, as you experienced it, including the four generations of AI successes

Lex Fridman (03:46.040)

that I've seen you talk about?

Lex Fridman (03:47.800)

Sure.

Michael I. Jordan (03:48.800)

Yeah, first of all, I much prefer Yann's metaphor.

Lex Fridman (03:54.040)

Miles Davis was a real explorer in jazz and he had a coherent story.

Lex Fridman (04:00.020)

So I think I have one, but it's not just the one you lived, it's the one you think about

Lex Fridman (04:04.320)

later.

Lex Fridman (04:05.320)

What the historian does is they look back and they revisit.

Michael I. Jordan (04:09.920)

I think what's happening right now is not AI, that was an intellectual aspiration that's

Michael I. Jordan (04:16.520)

still alive today as an aspiration.

Lex Fridman (04:18.640)

But I think this is akin to the development of chemical engineering from chemistry or

Michael I. Jordan (04:22.480)

electrical engineering from electromagnetism.

Lex Fridman (04:25.900)

So if you go back to the 30s or 40s, there wasn't yet chemical engineering.

Michael I. Jordan (04:31.040)

There was chemistry, there was fluid flow, there was mechanics and so on.

Lex Fridman (04:35.600)

But people pretty clearly viewed interesting goals to try to build factories that make

Michael I. Jordan (04:41.280)

chemicals products and do it viably, safely, make good ones, do it at scale.

Lex Fridman (04:48.060)

So people started to try to do that, of course, and some factories worked, some didn't, some

Michael I. Jordan (04:52.160)

were not viable, some exploded, but in parallel, developed a whole field called chemical engineering.

Michael I. Jordan (04:58.200)

Electrical engineering is a field, it's no bones about it, it has theoretical aspects

Michael I. Jordan (05:02.040)

to it, it has practical aspects.

Michael I. Jordan (05:04.720)

It's not just engineering, quote unquote, it's the real thing, real concepts are needed.

Michael I. Jordan (05:09.640)

Same thing with electrical engineering.

Michael I. Jordan (05:11.680)

There was Maxwell's equations, which in some sense were everything you know about electromagnetism,

Lex Fridman (05:16.620)

but you needed to figure out how to build circuits, how to build modules, how to put

Michael I. Jordan (05:19.120)

them together, how to bring electricity from one point to another safely and so on and

Lex Fridman (05:22.920)

so forth.

Lex Fridman (05:23.920)

So a whole field that developed called electrical engineering.

Michael I. Jordan (05:26.080)

I think that's what's happening right now, is that we have a proto field, which is statistics,

Michael I. Jordan (05:33.320)

more of the theoretical side of it, algorithmic side of computer science, that was enough

Lex Fridman (05:36.560)

to start to build things, but what things?

Michael I. Jordan (05:39.240)

Systems that bring value to human beings and use human data and mix in human decisions.

Michael I. Jordan (05:44.120)

The engineering side of that is all ad hoc.

Lex Fridman (05:47.620)

That's what's emerging.

Michael I. Jordan (05:48.620)

In fact, if you wanna call machine learning a field, I think that's what it is, that it's

Michael I. Jordan (05:51.560)

a proto form of engineering based on statistical and computational ideas of previous generations.

Lex Fridman (05:56.600)

But do you think there's something deeper about AI in his dreams and aspirations as

Lex Fridman (06:01.280)

compared to chemical engineering and electrical engineering?

Michael I. Jordan (06:03.840)

Well the dreams and aspirations maybe, but those are 500 years from now.

Michael I. Jordan (06:07.960)

I think that that's like the Greeks sitting there and saying, it would be neat to get

Michael I. Jordan (06:10.480)

to the moon someday.

Lex Fridman (06:12.920)

I think we have no clue how the brain does computation.

Michael I. Jordan (06:16.200)

We're just a clueless.

Michael I. Jordan (06:17.560)

We're even worse than the Greeks on most anything interesting scientifically of our era.

Lex Fridman (06:23.600)

Can you linger on that just for a moment because you stand not completely unique, but a little

Lex Fridman (06:29.100)

bit unique in the clarity of that.

Lex Fridman (06:31.400)

Can you elaborate your intuition of why we're, like where we stand in our understanding of

Lex Fridman (06:36.760)

the human brain?

Lex Fridman (06:37.760)

And a lot of people say, you know, scientists say we're not very far in understanding human

Lex Fridman (06:41.280)

brain, but you're like, you're saying we're in the dark here.

Michael I. Jordan (06:44.560)

Well, I know I'm not unique.

Michael I. Jordan (06:45.960)

I don't even think in the clarity, but if you talk to real neuroscientists that really

Michael I. Jordan (06:49.240)

study real synapses or real neurons, they agree, they agree.

Lex Fridman (06:53.480)

It's a hundreds of year task and they're building it up slowly and surely.

Lex Fridman (06:58.680)

What the signal is there is not clear.

Lex Fridman (07:00.920)

We think we have all of our metaphors.

Michael I. Jordan (07:02.700)

We think it's electrical, maybe it's chemical, it's a whole soup, it's ions and proteins

Lex Fridman (07:08.240)

and it's a cell.

Lex Fridman (07:09.240)

And that's even around like a single synapse.

Michael I. Jordan (07:11.080)

If you look at a electron micrograph of a single synapse, it's a city of its own.

Lex Fridman (07:15.920)

And that's one little thing on a dendritic tree, which is extremely complicated electrochemical

Lex Fridman (07:20.800)

thing.

Lex Fridman (07:22.000)

And it's doing these spikes and voltages are flying around and then proteins are taking

Lex Fridman (07:25.760)

that and taking it down into the DNA and who knows what.

Lex Fridman (07:29.440)

So it is the problem of the next few centuries.

Lex Fridman (07:31.760)

It is fantastic.

Lex Fridman (07:33.360)

But we have our metaphors about it.

Lex Fridman (07:34.940)

Is it an economic device?

Lex Fridman (07:36.160)

Is it like the immune system or is it like a layered set of, you know, arithmetic computations?

Lex Fridman (07:42.040)

We have all these metaphors and they're fun.

Lex Fridman (07:44.780)

But that's not real science per se.

Lex Fridman (07:48.120)

There is neuroscience.

Michael I. Jordan (07:49.120)

That's not neuroscience.

Lex Fridman (07:50.120)

All right.

Lex Fridman (07:51.120)

That's like the Greek speculating about how to get to the moon, fun, right?

Lex Fridman (07:55.380)

And I think that I like to say this fairly strongly because I think a lot of young people

Michael I. Jordan (07:59.040)

think we're on the verge because a lot of people who don't talk about it clearly let

Michael I. Jordan (08:03.440)

it be understood that, yes, we kind of, this is a brain inspired, we're kind of close,

Michael I. Jordan (08:07.520)

you know, breakthroughs are on the horizon.

Lex Fridman (08:10.280)

And that's scrupulous people sometimes who need money for their labs.

Michael I. Jordan (08:13.600)

That's what I'm saying, scrupulous, but people will oversell, I need money for my lab, I'm

Lex Fridman (08:18.680)

studying computational neuroscience, I'm going to oversell it.

Lex Fridman (08:23.880)

And so there's been too much of that.

Lex Fridman (08:25.200)

So I'll step into the gray area between metaphor and engineering with, I'm not sure if you're

Michael I. Jordan (08:32.040)

familiar with brain computer interfaces.

Lex Fridman (08:35.520)

So a company like Elon Musk has Neuralink that's working on putting electrodes into

Michael I. Jordan (08:42.240)

the brain and trying to be able to read, both read and send electrical signals.

Michael I. Jordan (08:46.520)

Just as you said, even the basic mechanism of communication in the brain is not something

Michael I. Jordan (08:54.320)

we understand.

Lex Fridman (08:55.320)

But do you hope without understanding the fundamental principles of how the brain works,

Lex Fridman (09:00.880)

we'll be able to do something interesting at that gray area of metaphor?

Lex Fridman (09:06.600)

It's not my area.

Lex Fridman (09:07.600)

So I hope in the sense, like anybody else hopes for some interesting things to happen

Michael I. Jordan (09:11.200)

from research, I would expect more something like Alzheimer's will get figured out from

Michael I. Jordan (09:15.600)

modern neuroscience.

Michael I. Jordan (09:16.600)

There's a lot of human suffering based on brain disease and we throw things like lithium

Michael I. Jordan (09:22.560)

at the brain, it kind of works, no one has a clue why.

Lex Fridman (09:25.900)

That's not quite true, but mostly we don't know.

Lex Fridman (09:28.240)

And that's even just about the biochemistry of the brain and how it leads to mood swings

Lex Fridman (09:31.940)

and so on.

Lex Fridman (09:33.120)

How thought emerges from that, we were really, really completely dim.

Lex Fridman (09:38.160)

So that you might want to hook up electrodes and try to do some signal processing on that

Lex Fridman (09:41.540)

and try to find patterns, fine, by all means, go for it.

Lex Fridman (09:45.640)

It's just not scientific at this point.

Lex Fridman (09:48.740)

So it's like kind of sitting in a satellite and watching the emissions from a city and

Michael I. Jordan (09:53.220)

trying to infer things about the microeconomy, even though you don't have microeconomic concepts.

Michael I. Jordan (09:57.680)

It's really that kind of thing.

Lex Fridman (09:59.200)

And so yes, can you find some signals that do something interesting or useful?

Lex Fridman (10:02.520)

Can you control a cursor or mouse with your brain?

Michael I. Jordan (10:06.640)

Yeah, absolutely, and then I can imagine business models based on that and even medical applications

Michael I. Jordan (10:13.240)

of that.

Lex Fridman (10:14.240)

But from there to understanding algorithms that allow us to really tie in deeply from

Michael I. Jordan (10:19.680)

the brain to computer, I just, no, I don't agree with Elon Musk.

Michael I. Jordan (10:22.580)

I don't think that's even, that's not for our generations, not even for the century.

Lex Fridman (10:26.580)

So just in hopes of getting you to dream, you've mentioned Kolmogorov and Turing might

Michael I. Jordan (10:33.580)

pop up, do you think that there might be breakthroughs that will get you to sit back in five, 10

Lex Fridman (10:41.120)

years and say, wow?

Michael I. Jordan (10:43.160)

Oh, I'm sure there will be, but I don't think that there'll be demos that impress me.

Michael I. Jordan (10:49.240)

I don't think that having a computer call a restaurant and pretend to be a human is

Lex Fridman (10:55.120)

a breakthrough.

Michael I. Jordan (10:56.120)

Right.

Lex Fridman (10:57.120)

And people, you know, some people present it as such.

Michael I. Jordan (10:59.840)

It's imitating human intelligence.

Lex Fridman (11:01.660)

It's even putting coughs in the thing to make a bit of a PR stunt.

Lex Fridman (11:07.400)

And so fine that the world runs on those things too.

Lex Fridman (11:11.440)

And I don't want to diminish all the hard work and engineering that goes behind things

Michael I. Jordan (11:14.940)

like that and the ultimate value to the human race.

Lex Fridman (11:17.760)

But that's not scientific understanding.

Lex Fridman (11:20.520)

And I know the people that work on these things, they are after scientific understanding.

Michael I. Jordan (11:23.880)

In the meantime, they've got to kind of, you know, the trains got to run and they got mouths

Michael I. Jordan (11:26.780)

to feed and they got things to do and there's nothing wrong with all that.

Lex Fridman (11:30.460)

I would call that though, just engineering.

Lex Fridman (11:32.560)

And I want to distinguish that between an engineering field, like electrical engineering

Lex Fridman (11:35.960)

and chemical engineering that originally emerged, that had real principles and you really know

Lex Fridman (11:39.360)

what you're doing and you had a little scientific understanding, maybe not even complete.

Lex Fridman (11:43.680)

So it became more predictable and it really gave value to human life because it was understood.

Lex Fridman (11:49.040)

And so we don't want to muddle too much these waters of, you know, what we're able to do

Lex Fridman (11:54.180)

versus what we really can't do in a way that's going to impress the next.

Lex Fridman (11:58.080)

So I don't need to be wowed, but I think that someone comes along in 20 years, a younger

Michael I. Jordan (12:02.520)

person who's absorbed all the technology and for them to be wowed, I think they have to

Michael I. Jordan (12:08.400)

be more deeply impressed.

Michael I. Jordan (12:09.400)

A young Kolmogorov would not be wowed by some of the stunts that you see right now coming

Michael I. Jordan (12:13.020)

from the big companies.

Michael I. Jordan (12:14.020)

The demos, but do you think the breakthroughs from Kolmogorov would be, and give this question

Michael I. Jordan (12:19.040)

a chance, do you think there'll be in the scientific fundamental principles arena or

Lex Fridman (12:24.120)

do you think it's possible to have fundamental breakthroughs in engineering?

Michael I. Jordan (12:28.400)

Meaning, you know, I would say some of the things that Elon Musk is working with SpaceX

Lex Fridman (12:33.200)

and then others sort of trying to revolutionize the fundamentals of engineering, of manufacturing,

Michael I. Jordan (12:39.840)

of saying, here's a problem we know how to do a demo of and actually taking it to scale.

Lex Fridman (12:44.480)

Yeah.

Lex Fridman (12:45.480)

So there's going to be all kinds of breakthroughs.

Lex Fridman (12:46.960)

I just don't like that terminology.

Michael I. Jordan (12:48.280)

I'm a scientist and I work on things day in and day out and things move along and eventually

Lex Fridman (12:52.000)

you say, wow, something happened, but I don't like that language very much.

Michael I. Jordan (12:56.400)

Also I don't like to prize theoretical breakthroughs over practical ones.

Michael I. Jordan (13:01.000)

I tend to be more of a theoretician and I think there's lots to do in that arena right

Michael I. Jordan (13:05.080)

now.

Lex Fridman (13:06.760)

And so I wouldn't point to the Kolmogorovs, I might point to the Edisons of the era and

Michael I. Jordan (13:09.800)

maybe Musk is a bit more like that.

Lex Fridman (13:11.840)

But you know, Musk, God bless him, also will say things about AI that he knows very little

Michael I. Jordan (13:17.440)

about and he leads people astray when he talks about things he doesn't know anything about.

Michael I. Jordan (13:23.840)

Trying to program a computer to understand natural language, to be involved in a dialogue

Michael I. Jordan (13:27.360)

we're having right now, that ain't going to happen in our lifetime.

Michael I. Jordan (13:30.460)

You could fake it, you can mimic, sort of take old sentences that humans use and retread

Michael I. Jordan (13:35.240)

them, but the deep understanding of language, no, it's not going to happen.

Lex Fridman (13:38.560)

And so from that, I hope you can perceive that the deeper, yet deeper kind of aspects

Lex Fridman (13:42.960)

and intelligence are not going to happen.

Lex Fridman (13:44.520)

Now will there be breakthroughs?

Michael I. Jordan (13:45.520)

No, I think that Google was a breakthrough, I think Amazon is a breakthrough, you know,

Michael I. Jordan (13:49.600)

I think Uber is a breakthrough, you know, that bring value to human beings at scale

Michael I. Jordan (13:53.280)

in new, brand new ways based on data flows and so on.

Michael I. Jordan (13:56.880)

A lot of these things are slightly broken because there's not kind of an engineering

Michael I. Jordan (14:01.260)

field that takes economic value in context of data and, you know, planetary scale and

Michael I. Jordan (14:06.680)

worries about all the externalities, the privacy, you know, we don't have that field so we don't

Michael I. Jordan (14:11.240)

think these things through very well.

Michael I. Jordan (14:13.200)

I see that as emerging and that will be, you know, looking back from 100 years, that will

Michael I. Jordan (14:17.560)

be a constituted breakthrough in this era, just like electrical engineering was a breakthrough

Michael I. Jordan (14:21.240)

in the early part of the last century and chemical engineering was a breakthrough.

Lex Fridman (14:24.560)

So the scale, the markets that you talk about and we'll get to will be seen as sort of breakthrough

Lex Fridman (14:30.360)

and we're in the very early days of really doing interesting stuff there and we'll get

Michael I. Jordan (14:34.360)

to that, but just taking a quick step back, can you give, kind of throw off the historian

Lex Fridman (14:40.920)

hat.

Michael I. Jordan (14:41.920)

I mean, you briefly said that the history of AI kind of mimics the history of chemical

Lex Fridman (14:47.760)

engineering, but...

Michael I. Jordan (14:49.240)

I keep saying machine learning.

Michael I. Jordan (14:50.360)

You keep wanting to say AI, just to let you know, I don't, you know, I resist that.

Michael I. Jordan (14:54.280)

I don't think this is about AI really was John McCarthy as almost a philosopher saying,

Lex Fridman (15:01.080)

wouldn't it be cool if we could put thought in a computer?

Michael I. Jordan (15:03.560)

If we could mimic the human capability to think or put intelligence in, in some sense

Lex Fridman (15:08.040)

into a computer.

Michael I. Jordan (15:09.960)

That's an interesting philosophical question and he wanted to make it more than philosophy.

Michael I. Jordan (15:13.560)

He wanted to actually write down a logical formula and algorithms that would do that.

Lex Fridman (15:17.340)

And that is a perfectly valid, reasonable thing to do.

Lex Fridman (15:20.180)

That's not what's happening in this era.

Lex Fridman (15:23.080)

So the reason I keep saying AI actually, and I'd love to hear what you think about it.

Lex Fridman (15:27.760)

Machine learning has a very particular set of methods and tools.

Michael I. Jordan (15:34.640)

Maybe your version of it is that mine doesn't, it's very, very open.

Lex Fridman (15:37.720)

It does optimization, it does sampling, it does...

Lex Fridman (15:40.160)

So systems that learn is what machine learning is.

Lex Fridman (15:42.920)

Systems that learn and make decisions.

Lex Fridman (15:44.560)

And make decisions.

Lex Fridman (15:45.560)

So it's not just pattern recognition and, you know, finding patterns, it's all about

Michael I. Jordan (15:49.080)

making decisions in real worlds and having close feedback loops.

Lex Fridman (15:52.560)

So something like symbolic AI, expert systems, reasoning systems, knowledge based representation,

Michael I. Jordan (15:58.400)

all of those kinds of things, search, does that neighbor fit into what you think of as

Lex Fridman (16:03.760)

machine learning?

Lex Fridman (16:04.760)

So I don't even like the word machine learning, I think that what the field you're talking

Michael I. Jordan (16:07.560)

about is all about making large collections of decisions under uncertainty by large collections

Michael I. Jordan (16:11.720)

of entities.

Lex Fridman (16:12.720)

Right?

Lex Fridman (16:13.720)

And there are principles for that, at that scale.

Michael I. Jordan (16:16.100)

You don't have to say the principles are for a single entity that's making decisions, single

Michael I. Jordan (16:19.040)

agent or single human.

Lex Fridman (16:20.560)

It really immediately goes to the network of decisions.

Lex Fridman (16:22.600)

Is a good word for that or no?

Lex Fridman (16:24.080)

No, there's no good words for any of this.

Michael I. Jordan (16:25.400)

That's kind of part of the problem.

Lex Fridman (16:27.240)

So we can continue the conversation to use AI for all that.

Michael I. Jordan (16:29.920)

I just want to kind of raise the flag here that this is not about, we don't know what

Lex Fridman (16:35.520)

intelligence is and real intelligence.

Michael I. Jordan (16:38.140)

We don't know much about abstraction and reasoning at the level of humans.

Lex Fridman (16:41.000)

We don't have a clue.

Michael I. Jordan (16:42.000)

We're not trying to build that because we don't have a clue.

Lex Fridman (16:44.720)

Eventually it may emerge.

Michael I. Jordan (16:45.720)

They'll make, I don't know if there'll be breakthroughs, but eventually we'll start

Lex Fridman (16:48.280)

to get glimmers of that.

Michael I. Jordan (16:50.160)

It's not what's happening right now.

Lex Fridman (16:51.480)

Okay.

Michael I. Jordan (16:52.480)

We're taking data.

Lex Fridman (16:53.480)

We're trying to make good decisions based on that.

Michael I. Jordan (16:54.560)

We're trying to scale.

Lex Fridman (16:55.560)

We're trying to economically viably, we're trying to build markets.

Michael I. Jordan (16:58.260)

We're trying to keep value at that scale and aspects of this will look intelligent.

Lex Fridman (17:04.680)

Computers were so dumb before, they will seem more intelligent.

Michael I. Jordan (17:08.120)

We will use that buzzword of intelligence so we can use it in that sense.

Lex Fridman (17:12.320)

So machine learning, you can scope it narrowly as just learning from data and pattern recognition.

Lex Fridman (17:17.960)

But when I talk about these topics, maybe data science is another word you could throw

Lex Fridman (17:22.140)

in the mix, it really is important that the decisions are as part of it.

Michael I. Jordan (17:26.880)

It's consequential decisions in the real world.

Lex Fridman (17:28.760)

Am I going to have a medical operation?

Lex Fridman (17:30.880)

Am I going to drive down the street?

Michael I. Jordan (17:33.480)

Things where there's scarcity, things that impact other human beings or other environments

Lex Fridman (17:38.240)

and so on.

Lex Fridman (17:39.400)

How do I do that based on data?

Lex Fridman (17:40.800)

How do I do that adaptively?

Lex Fridman (17:41.800)

How do I use computers to help those kinds of things go forward?

Michael I. Jordan (17:44.160)

Whatever you want to call that.

Lex Fridman (17:45.640)

So let's call it AI.

Michael I. Jordan (17:46.640)

Let's agree to call it AI, but let's not say that the goal of that is intelligence.

Michael I. Jordan (17:52.960)

The goal of that is really good working systems at planetary scale that we've never seen before.

Lex Fridman (17:56.640)

So reclaim the word AI from the Dartmouth conference from many decades ago of the dream

Lex Fridman (18:00.800)

of humans.

Michael I. Jordan (18:01.800)

I don't want to reclaim it.

Lex Fridman (18:02.800)

I want a new word.

Michael I. Jordan (18:03.800)

I think it was a bad choice.

Michael I. Jordan (18:04.800)

I mean, if you read one of my little things, the history was basically that McCarthy needed

Michael I. Jordan (18:09.820)

a new name because cybernetics already existed and he didn't like, no one really liked Norbert

Lex Fridman (18:14.800)

Wiener.

Michael I. Jordan (18:15.800)

Norbert Wiener was kind of an island to himself and he felt that he had encompassed all this

Lex Fridman (18:19.560)

and in some sense he did.

Michael I. Jordan (18:21.200)

You look at the language of cybernetics, it was everything we're talking about.

Michael I. Jordan (18:24.400)

It was control theory and signal processing and some notions of intelligence and closed

Michael I. Jordan (18:28.200)

feedback loops and data.

Lex Fridman (18:29.440)

It was all there.

Michael I. Jordan (18:30.960)

It's just not a word that lived on partly because of the maybe the personalities.

Lex Fridman (18:34.240)

But McCarthy needed a new word to say, I'm different from you.

Michael I. Jordan (18:36.720)

I'm not part of your show.

Lex Fridman (18:38.400)

I got my own.

Michael I. Jordan (18:40.080)

Invented this word and again, thinking forward about the movies that would be made about

Lex Fridman (18:46.240)

it, it was a great choice.

Lex Fridman (18:48.680)

But thinking forward about creating a sober academic and real world discipline, it was

Michael I. Jordan (18:52.000)

a terrible choice because it led to promises that are not true that we understand.

Michael I. Jordan (18:56.320)

We understand artificial perhaps, but we don't understand intelligence.

Michael I. Jordan (18:58.880)

It's a small tangent because you're one of the great personalities of machine learning,

Michael I. Jordan (19:03.360)

whatever the heck you call the field.

Lex Fridman (19:06.400)

Do you think science progresses by personalities or by the fundamental principles and theories

Lex Fridman (19:11.880)

and research that's outside of personalities?

Lex Fridman (19:15.080)

Both.

Lex Fridman (19:16.080)

And I wouldn't say there should be one kind of personality.

Michael I. Jordan (19:17.560)

I have mine and I have my preferences and I have a kind of network around me that feeds

Michael I. Jordan (19:23.200)

me and some of them agree with me and some of them disagree, but all kinds of personalities

Lex Fridman (19:26.680)

are needed.

Michael I. Jordan (19:28.480)

Right now, I think the personality that it's a little too exuberant, a little bit too ready

Lex Fridman (19:31.680)

to promise the moon is a little bit too much in ascendance.

Lex Fridman (19:35.840)

And I do think that there's some good to that.

Michael I. Jordan (19:38.160)

It certainly attracts lots of young people to our field, but a lot of those people come

Michael I. Jordan (19:41.580)

in with strong misconceptions and they have to then unlearn those and then find something

Lex Fridman (19:47.400)

to do.

Lex Fridman (19:48.880)

And so I think there's just got to be some multiple voices and I wasn't hearing enough

Lex Fridman (19:52.920)

of the more sober voice.

Lex Fridman (19:54.840)

So as a continuation of a fun tangent and speaking of vibrant personalities, what would

Lex Fridman (1:00:01.520)

Yeah.

Lex Fridman (1:00:02.520)

And you think technically speaking, it's possible to help.

Michael I. Jordan (1:00:07.800)

I don't know the answers, but it's a, it's a, it's a less anonymity, a little more locality,

Michael I. Jordan (1:00:13.840)

you know, worlds that you kind of enter in and you trust the people there in those worlds

Lex Fridman (1:00:17.340)

so that when you start having a discussion, you know, not only is that people are not

Michael I. Jordan (1:00:20.040)

going to hurt you, but it's not going to be a total waste of your time because there's

Michael I. Jordan (1:00:23.120)

a lot of wasting of time that, you know, a lot of us, I pulled out of Facebook early

Michael I. Jordan (1:00:26.640)

on cause it was clearly going to waste a lot of my time even though there was some value.

Lex Fridman (1:00:31.360)

And so, yeah, worlds that are somehow you enter in and you know what you're getting

Lex Fridman (1:00:34.600)

and it's kind of appeals to you and you might, new things might happen, but you kind of have

Lex Fridman (1:00:38.400)

some, some trust in that world.

Lex Fridman (1:00:40.820)

And there's some deep, interesting, complex psychological aspects around anonymity, how

Lex Fridman (1:00:46.520)

that changes human behavior that's quite dark.

Michael I. Jordan (1:00:49.960)

Quite dark.

Lex Fridman (1:00:50.960)

Yeah.

Michael I. Jordan (1:00:51.960)

I think a lot of us are, especially those of us who really loved the advent of technology.

Lex Fridman (1:00:55.440)

I love social networks when they came out.

Michael I. Jordan (1:00:56.760)

I was just, I didn't see any negatives there at all.

Lex Fridman (1:00:59.520)

But then I started seeing comment sections.

Michael I. Jordan (1:01:01.720)

I think it was maybe, you know, with the CNN or something.

Lex Fridman (1:01:04.760)

And I started to go, wow, this, this darkness I just did not know about and, and our technology

Michael I. Jordan (1:01:10.040)

is now amplifying it.

Lex Fridman (1:01:11.520)

So sorry for the big philosophical question, but on that topic, do you think human beings,

Michael I. Jordan (1:01:15.960)

cause you've also, out of all things, had a foot in psychology too, the, do you think

Lex Fridman (1:01:21.120)

human beings are fundamentally good?

Michael I. Jordan (1:01:23.800)

Like all of us have good intent that could be mind or is it depending on context and

Lex Fridman (1:01:32.240)

environment, everybody could be evil.

Lex Fridman (1:01:34.960)

So my answer is fundamentally good.

Lex Fridman (1:01:37.720)

But fundamentally limited.

Michael I. Jordan (1:01:39.240)

All of us have very, you know, blinkers on.

Lex Fridman (1:01:41.320)

We don't see the other person's pain that easily.

Michael I. Jordan (1:01:43.940)

We don't see the other person's point of view that easily.

Lex Fridman (1:01:46.680)

We're very much in our own head, in our own world.

Lex Fridman (1:01:49.880)

And on my good days, I think the technology could open us up to, you know, more perspectives

Lex Fridman (1:01:53.920)

and more less blinkered and more understanding, you know, a lot of wars in human history happened

Michael I. Jordan (1:01:58.560)

because of just ignorance.

Michael I. Jordan (1:01:59.560)

They didn't, they, they thought the other person was doing this while their person wasn't

Michael I. Jordan (1:02:02.600)

doing this.

Lex Fridman (1:02:03.600)

And we have a huge amounts of that.

Lex Fridman (1:02:05.440)

But in my lifetime, I've not seen technology really help in that way yet.

Lex Fridman (1:02:09.200)

And I do, I do, I do believe in that, but you know, no, I think fundamentally humans

Michael I. Jordan (1:02:13.600)

are good.

Michael I. Jordan (1:02:14.600)

The people suffer, people have grievances because you have grudges and those things

Michael I. Jordan (1:02:17.440)

cause them to do things they probably wouldn't want.

Lex Fridman (1:02:20.000)

They regret it often.

Lex Fridman (1:02:22.640)

So no, I, I think it's a, you know, part of the progress of technology is to indeed allow

Lex Fridman (1:02:28.080)

it to be a little easier to be the real good person you actually are.

Michael I. Jordan (1:02:31.160)

Well, but do you think individual human life or society could be modeled as an optimization

Lex Fridman (1:02:39.880)

problem?

Michael I. Jordan (1:02:40.880)

Not the way I think typically, I mean, that's, you're talking about one of the most complex

Michael I. Jordan (1:02:45.080)

phenomenon in the whole, you know, in all of which the individual human life or society

Michael I. Jordan (1:02:49.600)

as a whole.

Lex Fridman (1:02:50.600)

Both, both.

Michael I. Jordan (1:02:51.600)

I mean, individual human life is amazingly complex.

Lex Fridman (1:02:54.440)

And so you know, optimization is kind of just one branch of mathematics that talks about

Michael I. Jordan (1:02:58.960)

certain kinds of things.

Lex Fridman (1:02:59.960)

And it just feels way too limited for the complexity of such things.

Lex Fridman (1:03:04.520)

What properties of optimization problems do you think, so do you think most interesting

Michael I. Jordan (1:03:09.440)

problems that could be solved through optimization, what kind of properties does that surface

Lex Fridman (1:03:13.860)

have non convexity, convexity, linearity, all those kinds of things, saddle points?

Lex Fridman (1:03:19.680)

Well, so optimization is just one piece of mathematics.

Michael I. Jordan (1:03:22.160)

You know, there's like, you just, even in our era, we're aware that say sampling is

Lex Fridman (1:03:27.480)

coming up, examples of something coming up with a distribution.

Lex Fridman (1:03:31.520)

What's optimization?

Lex Fridman (1:03:32.520)

What's sampling?

Michael I. Jordan (1:03:33.520)

Well, they, you can, if you're a kind of a certain kind of mathematician, you can try

Lex Fridman (1:03:35.920)

to blend them and make them seem to be sort of the same thing.

Lex Fridman (1:03:38.680)

But optimization is roughly speaking, trying to find a point that, a single point that

Lex Fridman (1:03:44.160)

is the optimum of a criterion function of some kind.

Lex Fridman (1:03:48.740)

And sampling is trying to, from that same surface, treat that as a distribution or density

Lex Fridman (1:03:53.940)

and find points that have high density.

Lex Fridman (1:03:56.920)

So I want the entire distribution in a sampling paradigm and I want the, you know, the single

Lex Fridman (1:04:03.480)

point, that's the best point in the optimization paradigm.

Michael I. Jordan (1:04:07.640)

Now if you were optimizing in the space of probability measures, the output of that could

Lex Fridman (1:04:11.880)

be a whole probability distribution.

Lex Fridman (1:04:13.080)

So you can start to make these things the same.

Lex Fridman (1:04:15.560)

But in mathematics, if you go too high up that kind of abstraction hierarchy, you start

Michael I. Jordan (1:04:18.400)

to lose the, you know, the ability to do the interesting theorems.

Lex Fridman (1:04:22.900)

So you kind of don't try that.

Michael I. Jordan (1:04:23.900)

You don't try to overly over abstract.

Lex Fridman (1:04:26.960)

So as a small tangent, what kind of worldview do you find more appealing?

Lex Fridman (1:04:31.540)

One that is deterministic or stochastic?

Lex Fridman (1:04:35.080)

Well, that's easy.

Michael I. Jordan (1:04:36.880)

I mean, I'm a statistician.

Lex Fridman (1:04:38.160)

You know, the world is highly stochastic.

Lex Fridman (1:04:40.400)

I don't know what's going to happen in the next five minutes, right?

Lex Fridman (1:04:42.360)

Because what you're going to ask, what we're going to do, what I'll say.

Michael I. Jordan (1:04:44.360)

Due to the uncertainty.

Lex Fridman (1:04:45.360)

Due to the...

Michael I. Jordan (1:04:46.360)

Massive uncertainty.

Lex Fridman (1:04:47.360)

Yeah.

Michael I. Jordan (1:04:48.360)

You know, massive uncertainty.

Lex Fridman (1:04:49.360)

And so the best I can do is have come rough sense or probability distribution on things

Lex Fridman (1:04:53.080)

and somehow use that in my reasoning about what to do now.

Lex Fridman (1:04:58.280)

So how does the distributed at scale when you have multi agent systems look like?

Lex Fridman (1:05:07.080)

So optimization can optimize sort of, it makes a lot more sense, sort of at least from my

Michael I. Jordan (1:05:13.760)

from robotics perspective, for a single robot, for a single agent, trying to optimize some

Michael I. Jordan (1:05:18.240)

objective function.

Michael I. Jordan (1:05:21.040)

When you start to enter the real world, this game theoretic concept starts popping up.

Lex Fridman (1:05:27.080)

That's how do you see optimization in this?

Lex Fridman (1:05:30.400)

Because you've talked about markets in a scale.

Lex Fridman (1:05:32.720)

What does that look like?

Lex Fridman (1:05:33.720)

Do you see it as optimization?

Lex Fridman (1:05:34.720)

Do you see it as sampling?

Lex Fridman (1:05:36.120)

Do you see like, how should you mark?

Michael I. Jordan (1:05:38.280)

These all blend together.

Lex Fridman (1:05:39.280)

And a system designer thinking about how to build an incentivized system will have a blend

Michael I. Jordan (1:05:44.120)

of all these things.

Michael I. Jordan (1:05:45.120)

So, you know, a particle in a potential well is optimizing a functional called a Lagrangian,

Lex Fridman (1:05:49.800)

right?

Lex Fridman (1:05:50.800)

The particle doesn't know that.

Michael I. Jordan (1:05:51.800)

There's no algorithm running that does that.

Lex Fridman (1:05:54.640)

It just happens.

Lex Fridman (1:05:55.640)

And so it's a description mathematically of something that helps us understand as analysts

Lex Fridman (1:05:59.160)

what's happening, right?

Lex Fridman (1:06:00.840)

And so the same thing will happen when we talk about, you know, mixtures of humans and

Michael I. Jordan (1:06:03.520)

computers and markets and so on and so forth, there'll be certain principles that allow

Michael I. Jordan (1:06:07.080)

us to understand what's happening, whether or not the actual algorithms are being used

Lex Fridman (1:06:10.320)

by any sense is not clear.

Michael I. Jordan (1:06:13.000)

Now at some point, I may have set up a multi agent or market kind of system.

Lex Fridman (1:06:19.080)

And I'm now thinking about an individual agent in that system.

Lex Fridman (1:06:22.440)

And they're asked to do some task and they're incentivized in some way, they get certain

Lex Fridman (1:06:25.200)

signals and they have some utility.

Lex Fridman (1:06:28.160)

What they will do at that point is they just won't know the answer, they may have to optimize

Lex Fridman (1:06:31.560)

to find an answer.

Michael I. Jordan (1:06:32.560)

Okay, so an artist could be embedded inside of an overall market.

Lex Fridman (1:06:36.920)

You know, and game theory is very, very broad.

Michael I. Jordan (1:06:39.880)

It is often studied very narrowly for certain kinds of problems.

Lex Fridman (1:06:44.020)

But it's roughly speaking, this is just the, I don't know what you're going to do.

Lex Fridman (1:06:47.920)

So I kind of anticipate that a little bit, and you anticipate what I'm anticipating.

Lex Fridman (1:06:51.560)

And we kind of go back and forth in our own minds.

Michael I. Jordan (1:06:53.360)

We run kind of thought experiments.

Michael I. Jordan (1:06:55.480)

You've talked about this interesting point in terms of game theory, you know, most optimization

Michael I. Jordan (1:07:00.420)

problems really hate saddle points, maybe you can describe what saddle points are.

Lex Fridman (1:07:04.840)

But I've heard you kind of mentioned that there's a there's a branch of optimization

Michael I. Jordan (1:07:09.560)

that you could try to explicitly look for saddle points as a good thing.

Lex Fridman (1:07:14.720)

Oh, not optimization.

Michael I. Jordan (1:07:15.840)

That's just game theory that that so there's all kinds of different equilibria in game

Lex Fridman (1:07:19.740)

theory.

Lex Fridman (1:07:20.740)

And some of them are highly explanatory behavior.

Lex Fridman (1:07:23.220)

They're not attempting to be algorithmic.

Michael I. Jordan (1:07:24.760)

They're just trying to say, if you happen to be at this equilibrium, you would see certain

Lex Fridman (1:07:29.080)

kind of behavior.

Lex Fridman (1:07:30.080)

And we see that in real life.

Michael I. Jordan (1:07:31.080)

That's what an economist wants to do, especially behavioral economists in continuous differential

Michael I. Jordan (1:07:39.420)

game theory, you're in continuous spaces, a some of the simplest equilibria are saddle

Lex Fridman (1:07:44.020)

points and Nash equilibrium as a saddle point.

Michael I. Jordan (1:07:46.400)

It's a special kind of saddle point.

Lex Fridman (1:07:48.440)

So classically, in game theory, you were trying to find Nash equilibria and an algorithmic

Michael I. Jordan (1:07:53.560)

game theory, you're trying to find algorithms that would find them.

Lex Fridman (1:07:56.400)

And so you're trying to find saddle points.

Michael I. Jordan (1:07:57.760)

I mean, so that's literally what you're trying to do.

Lex Fridman (1:08:00.720)

But you know, any economist knows that Nash equilibria have their limitations.

Michael I. Jordan (1:08:04.160)

They are definitely not that explanatory in many situations.

Lex Fridman (1:08:08.200)

They're not what you really want.

Michael I. Jordan (1:08:10.360)

There's other kind of equilibria.

Lex Fridman (1:08:12.180)

And there's names associated with these because they came from history with certain people

Michael I. Jordan (1:08:15.440)

working on them, but there will be new ones emerging.

Lex Fridman (1:08:18.080)

So you know, one example is a Stackelberg equilibrium.

Lex Fridman (1:08:21.200)

So you know, Nash, you and I are both playing this game against each other or for each other,

Michael I. Jordan (1:08:25.800)

maybe it's cooperative, and we're both going to think it through and then we're going to

Michael I. Jordan (1:08:29.000)

decide and we're going to do our thing simultaneously.

Lex Fridman (1:08:32.520)

You know, in a Stackelberg, no, I'm going to be the first mover.

Michael I. Jordan (1:08:34.640)

I'm going to make a move.

Lex Fridman (1:08:35.880)

You're going to look at my move and then you're going to make yours.

Michael I. Jordan (1:08:38.400)

Now since I know you're going to look at my move, I anticipate what you're going to do.

Lex Fridman (1:08:42.180)

And so I don't do something stupid, but then I know that you are also anticipating me.

Lex Fridman (1:08:46.960)

So we're kind of going back and forth on why, but there is then a first mover thing.

Lex Fridman (1:08:51.800)

And so those are different equilibria, right?

Lex Fridman (1:08:54.920)

And so just mathematically, yeah, these things have certain topologies and certain shapes

Lex Fridman (1:08:59.220)

that are like, what's it, algorithmically or dynamically, how do you move towards them?

Lex Fridman (1:09:02.840)

How do you move away from things?

Michael I. Jordan (1:09:05.820)

You know, so some of these questions have answers, they've been studied, others do not.

Lex Fridman (1:09:09.500)

And especially if it becomes stochastic, especially if there's large numbers of decentralized

Michael I. Jordan (1:09:13.920)

things, there's just, you know, young people get in this field who kind of think it's all

Michael I. Jordan (1:09:17.440)

done because we have, you know, TensorFlow.

Michael I. Jordan (1:09:19.520)

Well, no, these are all open problems and they're really important and interesting.

Lex Fridman (1:09:23.680)

And it's about strategic settings.

Lex Fridman (1:09:25.140)

How do I collect data?

Lex Fridman (1:09:26.640)

Suppose I don't know what you're going to do because I don't know you very well, right?

Lex Fridman (1:09:29.280)

Well, I got to collect data about you.

Lex Fridman (1:09:31.180)

So maybe I want to push you into a part of the space where I don't know much about you

Lex Fridman (1:09:34.280)

so I can get data.

Michael I. Jordan (1:09:35.280)

Cause, and then later I'll realize that you'll never, you'll never go there because of the

Lex Fridman (1:09:38.960)

way the game is set up.

Michael I. Jordan (1:09:39.960)

You know, that's part of the overall, you know, data analysis context is that.

Michael I. Jordan (1:09:44.080)

Even the game of poker is fascinating space, whenever there's any uncertainty, a lack of

Michael I. Jordan (1:09:47.840)

information, it's a super exciting space.

Lex Fridman (1:09:52.560)

Just to linger on optimization for a second.

Lex Fridman (1:09:55.360)

So when we look at deep learning, it's essentially minimization of a complicated loss function.

Lex Fridman (1:10:01.600)

So is there something insightful or hopeful that you see in the kinds of function surface

Lex Fridman (1:10:07.400)

that loss functions, the deep learning and in the real world is trying to optimize over?

Lex Fridman (1:10:13.800)

Is there something interesting as it's just the usual kind of problems of optimization?

Michael I. Jordan (1:10:20.040)

I think from an optimization point of view, that surface, first of all, it's pretty smooth.

Lex Fridman (1:10:25.600)

And secondly, if there's over, if it's over parameterized, there's kind of lots of paths

Michael I. Jordan (1:10:29.120)

down to reasonable Optima.

Lex Fridman (1:10:31.540)

And so kind of the getting downhill to the, to an optimum is viewed as not as hard as

Michael I. Jordan (1:10:35.680)

you might've expected in high dimensions.

Lex Fridman (1:10:39.980)

The fact that some Optima tend to be really good ones and others not so good.

Lex Fridman (1:10:43.200)

And you tend to, it's not, sometimes you find the good ones is sort of still needs explanation.

Lex Fridman (1:10:48.080)

Yeah.

Michael I. Jordan (1:10:49.080)

But, but the particular surface is coming from the particular generation of neural nets.

Lex Fridman (1:10:53.560)

I kind of suspect those will, those will change in 10 years.

Michael I. Jordan (1:10:56.880)

It will not be exactly those surfaces.

Michael I. Jordan (1:10:58.360)

There'll be some others that are an optimization theory will help contribute to why other surfaces

Michael I. Jordan (1:11:02.500)

or why other algorithms.

Michael I. Jordan (1:11:05.640)

Years of arithmetic operations with a little bit of nonlinearity, that's not, that didn't

Michael I. Jordan (1:11:09.840)

come from neuroscience per se.

Michael I. Jordan (1:11:10.960)

I mean, maybe in the minds of some of the people working on it, they were thinking about

Michael I. Jordan (1:11:13.920)

brains, but they were arithmetic circuits in all kinds of fields, computer science control

Lex Fridman (1:11:19.040)

theory and so on.

Lex Fridman (1:11:20.640)

And that layers of these could transform things in certain ways.

Lex Fridman (1:11:23.480)

And that if it's smooth, maybe you could find parameter values is a sort of big discovery

Michael I. Jordan (1:11:32.000)

that it's working, it's able to work at this scale.

Lex Fridman (1:11:35.000)

But I don't think that we're stuck with that and we're, we're certainly not stuck with

Michael I. Jordan (1:11:39.840)

that cause we're understanding the brain.

Lex Fridman (1:11:42.120)

So in terms of on the algorithm side sort of gradient descent, do you think we're stuck

Lex Fridman (1:11:46.360)

with gradient descent as a variance of it?

Lex Fridman (1:11:49.360)

What variance do you find interesting or do you think there'll be something else invented

Lex Fridman (1:11:53.600)

that is able to walk all over these optimization spaces in more interesting ways?

Lex Fridman (1:11:59.720)

So there's a co design of the surface and the, or the architecture and the algorithm.

Lex Fridman (1:12:04.700)

So if you just ask if we stay with the kind of architectures that we have now and not

Michael I. Jordan (1:12:08.080)

just neural nets, but you know, phase retrieval architectures or matrix completion architectures

Lex Fridman (1:12:13.080)

and so on.

Michael I. Jordan (1:12:15.080)

You know, I think we've kind of come to a place where yeah, a stochastic gradient algorithms

Michael I. Jordan (1:12:19.560)

are dominant and there are versions that are a little better than others.

Lex Fridman (1:12:25.840)

They have more guarantees, they're more robust and so on.

Lex Fridman (1:12:29.160)

And there's ongoing research to kind of figure out which is the best arm for which situation.

Lex Fridman (1:12:34.260)

But I think that that'll start to co evolve, that that'll put pressure on the actual architecture.

Lex Fridman (1:12:37.880)

And so we shouldn't do it in this particular way, we should do it in a different way because

Lex Fridman (1:12:40.800)

this other algorithm is now available if you do it in a different way.

Lex Fridman (1:12:45.340)

So that I can't really anticipate that co evolution process, but you know, gradients

Lex Fridman (1:12:51.600)

are amazing mathematical objects.

Michael I. Jordan (1:12:54.480)

They have a lot of people who start to study them more deeply mathematically are kind of

Lex Fridman (1:13:01.120)

shocked about what they are and what they can do.

Michael I. Jordan (1:13:05.160)

Think about it this way, suppose that I tell you if you move along the x axis, you go uphill

Michael I. Jordan (1:13:11.040)

in some objective by three units, whereas if you move along the y axis, you go uphill

Lex Fridman (1:13:15.920)

by seven units, right?

Lex Fridman (1:13:18.000)

Now I'm going to only allow you to move a certain unit distance, right?

Lex Fridman (1:13:22.440)

What are you going to do?

Michael I. Jordan (1:13:23.440)

Well, most people will say that I'm going to go along the y axis, I'm getting the biggest

Michael I. Jordan (1:13:27.240)

bang for my buck, you know, and my buck is only one unit, so I'm going to put all of

Lex Fridman (1:13:31.120)

it in the y axis, right?

Lex Fridman (1:13:33.960)

And why should I even take any of my strength, my step size and put any of it in the x axis

Lex Fridman (1:13:39.320)

because I'm getting less bang for my buck.

Michael I. Jordan (1:13:41.480)

That seems like a completely clear argument and it's wrong because the gradient direction

Lex Fridman (1:13:47.480)

is not to go along the y axis, it's to take a little bit of the x axis.

Lex Fridman (1:13:51.780)

And to understand that, you have to know some math and so even a trivial so called operator

Michael I. Jordan (1:13:59.200)

like gradient is not trivial and so, you know, exploiting its properties is still very important.

Michael I. Jordan (1:14:04.020)

Now we know that just pervading descent has got all kinds of problems, it gets stuck in

Lex Fridman (1:14:06.840)

many ways and it had never, you know, good dimension dependence and so on.

Lex Fridman (1:14:10.960)

So my own line of work recently has been about what kinds of stochasticity, how can we get

Michael I. Jordan (1:14:15.960)

dimension dependence, how can we do the theory of that and we've come up pretty favorable

Michael I. Jordan (1:14:20.200)

results with certain kinds of stochasticity.

Lex Fridman (1:14:22.720)

We have sufficient conditions generally.

Michael I. Jordan (1:14:25.000)

We know if you do this, we will give you a good guarantee.

Michael I. Jordan (1:14:28.760)

We don't have necessary conditions that it must be done a certain way in general.

Lex Fridman (1:14:32.280)

So stochasticity, how much randomness to inject into the walking along the gradient?

Lex Fridman (1:14:38.200)

And what kind of randomness?

Lex Fridman (1:14:40.000)

Why is randomness good in this process?

Lex Fridman (1:14:42.240)

Why is stochasticity good?

Michael I. Jordan (1:14:44.240)

Yeah, so I can give you simple answers but in some sense again, it's kind of amazing.

Michael I. Jordan (1:14:49.320)

Stochasticity just, you know, particular features of a surface that could have hurt you if you

Michael I. Jordan (1:14:55.600)

were doing one thing deterministically won't hurt you because by chance, there's very little

Lex Fridman (1:15:02.080)

chance that you would get hurt.

Lex Fridman (1:15:04.800)

So here stochasticity, it just kind of saves you from some of the particular features of

Lex Fridman (1:15:12.840)

surfaces.

Michael I. Jordan (1:15:13.840)

In fact, if you think about surfaces that are discontinuous in our first derivative,

Michael I. Jordan (1:15:19.400)

like an absolute value function, you will go down and hit that point where there's nondifferentiability.

Lex Fridman (1:15:25.400)

And if you're running a deterministic algorithm at that point, you can really do something

Lex Fridman (1:15:28.520)

bad.

Michael I. Jordan (1:15:29.520)

Whereas stochasticity just means it's pretty unlikely that's going to happen, that you're

Lex Fridman (1:15:32.960)

going to hit that point.

Lex Fridman (1:15:35.720)

So it's again, nontrivial to analyze but especially in higher dimensions, also stochasticity,

Michael I. Jordan (1:15:41.860)

our intuition isn't very good about it but it has properties that kind of are very appealing

Michael I. Jordan (1:15:45.440)

in high dimensions for a lot of large number of reasons.

Lex Fridman (1:15:49.200)

So it's all part of the mathematics to kind of, that's what's fun to work in the field

Michael I. Jordan (1:15:52.520)

is that you get to try to understand this mathematics.

Lex Fridman (1:15:57.040)

But long story short, you know, partly empirically, it was discovered stochastic gradient is very

Michael I. Jordan (1:16:01.200)

effective and theory kind of followed, I'd say, that but I don't see that we're getting

Lex Fridman (1:16:06.600)

clearly out of that.

Lex Fridman (1:16:09.120)

What's the most beautiful, mysterious, a profound idea to you in optimization?

Lex Fridman (1:16:15.560)

I don't know the most.

Lex Fridman (1:16:17.360)

But let me just say that Nesterov's work on Nesterov acceleration to me is pretty surprising

Lex Fridman (1:16:23.600)

and pretty deep.

Lex Fridman (1:16:26.280)

Can you elaborate?

Michael I. Jordan (1:16:27.280)

Well Nesterov acceleration is just that, suppose that we are going to use gradients

Michael I. Jordan (1:16:32.240)

to move around in a space.

Lex Fridman (1:16:33.240)

For the reasons I've alluded to, they're nice directions to move.

Lex Fridman (1:16:37.280)

And suppose that I tell you that you're only allowed to use gradients, you're not going

Michael I. Jordan (1:16:40.520)

to be allowed to use this local person that can only sense kind of the change in the surface.

Lex Fridman (1:16:47.440)

But I'm going to give you kind of a computer that's able to store all your previous gradients.

Lex Fridman (1:16:50.920)

And so you start to learn some something about the surface.

Lex Fridman (1:16:55.020)

And I'm going to restrict you to maybe move in the direction of like a linear span of

Lex Fridman (1:16:58.620)

all the gradients.

Lex Fridman (1:16:59.620)

So you can't kind of just move in some arbitrary direction, right?

Lex Fridman (1:17:02.880)

So now we have a well defined mathematical complexity model.

Michael I. Jordan (1:17:05.720)

There's certain classes of algorithms that can do that and others that can't.

Lex Fridman (1:17:09.320)

And we can ask for certain kinds of surfaces, how fast can you get down to the optimum?

Lex Fridman (1:17:13.800)

So there's answers to these.

Lex Fridman (1:17:14.960)

So for a smooth convex function, there's an answer, which is one over the number of steps

Michael I. Jordan (1:17:21.040)

squared.

Lex Fridman (1:17:22.400)

You will be within a ball of that size after k steps.

Michael I. Jordan (1:17:29.120)

Gradient descent in particular has a slower rate, it's one over k.

Lex Fridman (1:17:35.420)

So you could ask, is gradient descent actually, even though we know it's a good algorithm,

Lex Fridman (1:17:38.960)

is it the best algorithm?

Lex Fridman (1:17:39.960)

And the answer is no.

Michael I. Jordan (1:17:41.960)

Well, not clear yet, because one over k squared is a lower bound.

Lex Fridman (1:17:47.420)

That's probably the best you can do.

Lex Fridman (1:17:49.960)

Gradient is one over k, but is there something better?

Lex Fridman (1:17:52.740)

And so I think as a surprise to most, Nesterov discovered a new algorithm that has got two

Michael I. Jordan (1:17:59.280)

pieces to it.

Lex Fridman (1:18:00.280)

It's two gradients and puts those together in a certain kind of obscure way.

Lex Fridman (1:18:06.640)

And the thing doesn't even move downhill all the time.

Lex Fridman (1:18:09.280)

It sometimes goes back uphill.

Lex Fridman (1:18:10.760)

And if you're a physicist, that kind of makes some sense.

Michael I. Jordan (1:18:13.160)

You're building up some momentum and that is kind of the right intuition, but that intuition

Michael I. Jordan (1:18:17.720)

is not enough to understand kind of how to do it and why it works.

Lex Fridman (1:18:22.460)

But it does.

Michael I. Jordan (1:18:23.460)

It achieves one over k squared and it has a mathematical structure and it's still kind

Michael I. Jordan (1:18:27.520)

of to this day, a lot of us are writing papers and trying to explore that and understand

Michael I. Jordan (1:18:31.160)

it.

Lex Fridman (1:18:32.560)

So there are lots of cool ideas and optimization, but just kind of using gradients, I think

Michael I. Jordan (1:18:36.680)

is number one that goes back, you know, 150 years.

Lex Fridman (1:18:40.760)

And then Nesterov, I think has made a major contribution with this idea.

Lex Fridman (1:18:43.580)

So like you said, gradients themselves are in some sense, mysterious.

Lex Fridman (1:18:47.840)

They're not as trivial as...

Michael I. Jordan (1:18:50.440)

Not as trivial.

Lex Fridman (1:18:52.040)

Coordinate descent is more of a trivial one.

Michael I. Jordan (1:18:54.080)

You just pick one of the coordinates.

Lex Fridman (1:18:55.080)

That's how we think.

Michael I. Jordan (1:18:56.080)

That's how our human mind thinks.

Lex Fridman (1:18:57.080)

That's how our human minds think.

Lex Fridman (1:18:58.200)

And gradients are not that easy for our human mind to grapple with.

Lex Fridman (1:19:03.280)

An absurd question, but what is statistics?

Lex Fridman (1:19:08.600)

So here it's a little bit, it's somewhere between math and science and technology.

Lex Fridman (1:19:12.160)

It's somewhere in that convex hole.

Lex Fridman (1:19:13.420)

So it's a set of principles that allow you to make inferences that have got some reason

Michael I. Jordan (1:19:17.720)

to be believed and also principles that allow you to make decisions where you can have some

Michael I. Jordan (1:19:22.640)

reason to believe you're not going to make errors.

Lex Fridman (1:19:25.100)

So all of that requires some assumptions about what do you mean by an error?

Lex Fridman (1:19:27.680)

What do you mean by the probabilities?

Lex Fridman (1:19:31.420)

But after you start making some of those assumptions, you're led to conclusions that, yes, I can

Michael I. Jordan (1:19:38.080)

guarantee that if you do this in this way, your probability of making an error will be

Lex Fridman (1:19:42.080)

small.

Michael I. Jordan (1:19:43.600)

Your probability of continuing to not make errors over time will be small.

Lex Fridman (1:19:47.880)

And the probability that you found something that's real will be small, will be high.

Lex Fridman (1:19:52.280)

So decision making is a big part of that.

Lex Fridman (1:19:54.640)

Decision making is a big part.

Michael I. Jordan (1:19:55.640)

Yeah.

Lex Fridman (1:19:56.640)

So statistics, short history was that, it goes back as a formal discipline, 250 years

Michael I. Jordan (1:20:03.600)

or so.

Michael I. Jordan (1:20:04.960)

It was called inverse probability because around that era, probability was developed

Michael I. Jordan (1:20:09.280)

sort of especially to explain gambling situations.

Lex Fridman (1:20:12.000)

Of course, interesting.

Lex Fridman (1:20:15.480)

So you would say, well, given the state of nature is this, there's a certain roulette

Lex Fridman (1:20:18.880)

board that has a certain mechanism and what kind of outcomes do I expect to see?

Lex Fridman (1:20:23.680)

And especially if I do things long amounts of time, what outcomes will I see?

Lex Fridman (1:20:27.440)

And the physicists started to pay attention to this.

Lex Fridman (1:20:30.640)

And then people said, well, let's turn the problem around.

Lex Fridman (1:20:33.500)

What if I saw certain outcomes, could I infer what the underlying mechanism was?

Michael I. Jordan (1:20:37.480)

That's an inverse problem.

Lex Fridman (1:20:38.480)

And in fact, for quite a while, statistics was called inverse probability.

Michael I. Jordan (1:20:41.640)

That was the name of the field.

Lex Fridman (1:20:44.060)

And I believe that it was Laplace who was working in Napoleon's government who needed

Michael I. Jordan (1:20:50.600)

to do a census of France, learn about the people there.

Lex Fridman (1:20:54.280)

So he went and gathered data and he analyzed that data to determine policy and said, well,

Michael I. Jordan (1:21:01.240)

let's call this field that does this kind of thing statistics because the word state

Lex Fridman (1:21:06.760)

is in there.

Michael I. Jordan (1:21:07.760)

In French, that's etat, but it's the study of data for the state.

Lex Fridman (1:21:12.360)

So anyway, that caught on and it's been called statistics ever since.

Lex Fridman (1:21:18.640)

But by the time it got formalized, it was sort of in the 30s.

Lex Fridman (1:21:23.280)

And around that time, there was game theory and decision theory developed nearby.

Michael I. Jordan (1:21:28.560)

People in that era didn't think of themselves as either computer science or statistics or

Lex Fridman (1:21:31.640)

control or econ.

Michael I. Jordan (1:21:32.640)

They were all the above.

Lex Fridman (1:21:34.540)

And so Von Neumann is developing game theory, but also thinking of that as decision theory.

Michael I. Jordan (1:21:39.320)

Wald is an econometrician developing decision theory and then turning that into statistics.

Lex Fridman (1:21:45.120)

And so it's all about, here's not just data and you analyze it, here's a loss function.

Michael I. Jordan (1:21:50.160)

Here's what you care about.

Lex Fridman (1:21:51.160)

Here's the question you're trying to ask.

Michael I. Jordan (1:21:53.080)

Here is a probability model and here's the risk you will face if you make certain decisions.

Lex Fridman (1:21:59.440)

And to this day, in most advanced statistical curricula, you teach decision theory as the

Michael I. Jordan (1:22:04.040)

starting point and then it branches out into the two branches of Bayesian and frequentist.

Lex Fridman (1:22:08.500)

But that's all about decisions.

Michael I. Jordan (1:22:11.840)

In statistics, what is the most beautiful, mysterious, maybe surprising idea that you've

Lex Fridman (1:22:19.040)

come across?

Michael I. Jordan (1:22:20.040)

Yeah, good question.

Lex Fridman (1:22:21.040)

I mean, there's a bunch of surprising ones.

Michael I. Jordan (1:22:27.640)

There's something that's way too technical for this thing, but something called James

Michael I. Jordan (1:22:30.320)

Stein estimation, which is kind of surprising and really takes time to wrap your head around.

Lex Fridman (1:22:36.040)

Can you try to maybe...

Lex Fridman (1:22:37.040)

I think I don't want to even want to try.

Michael I. Jordan (1:22:39.120)

Let me just say a colleague at Steven Stigler at University of Chicago wrote a really beautiful

Lex Fridman (1:22:44.200)

paper on James Stein estimation, which helps to...

Michael I. Jordan (1:22:47.200)

It's views a paradox.

Michael I. Jordan (1:22:48.600)

It kind of defeats the mind's attempts to understand it, but you can and Steve has a

Michael I. Jordan (1:22:52.240)

nice perspective on that.

Lex Fridman (1:22:56.560)

So one of the troubles with statistics is that it's like in physics that are in quantum

Michael I. Jordan (1:23:00.320)

physics, you have multiple interpretations.

Michael I. Jordan (1:23:02.520)

There's a wave and particle duality in physics and you get used to that over time, but it

Michael I. Jordan (1:23:07.600)

still kind of haunts you that you don't really quite understand the relationship.

Lex Fridman (1:23:11.680)

The electron's a wave and electron's a particle.

Michael I. Jordan (1:23:15.840)

Well the same thing happens here.

Lex Fridman (1:23:16.840)

There's Bayesian ways of thinking and frequentist, and they are different.

Michael I. Jordan (1:23:21.320)

They sometimes become sort of the same in practice, but they are physically different.

Lex Fridman (1:23:25.000)

And then in some practice, they are not the same at all.

Michael I. Jordan (1:23:27.640)

They give you rather different answers.

Lex Fridman (1:23:30.480)

And so it is very much like wave and particle duality, and that is something that you have

Michael I. Jordan (1:23:33.860)

to kind of get used to in the field.

Lex Fridman (1:23:35.840)

Can you define Bayesian and frequentist?

Michael I. Jordan (1:23:37.720)

Yeah in decision theory you can make, I have a video that people could see.

Michael I. Jordan (1:23:41.320)

It's called are you a Bayesian or a frequentist and kind of help try to make it really clear.

Michael I. Jordan (1:23:46.040)

It comes from decision theory.

Lex Fridman (1:23:47.160)

So you know, decision theory, you're talking about loss functions, which are a function

Michael I. Jordan (1:23:51.920)

of data X and parameter theta.

Lex Fridman (1:23:54.760)

They're a function of two arguments.

Michael I. Jordan (1:23:57.080)

Okay.

Lex Fridman (1:23:58.080)

Neither one of those arguments is known.

Michael I. Jordan (1:23:59.880)

You don't know the data a priori.

Lex Fridman (1:24:01.640)

It's random and the parameters unknown.

Michael I. Jordan (1:24:03.760)

All right.

Lex Fridman (1:24:04.760)

So you have a function of two things you don't know, and you're trying to say, I want that

Michael I. Jordan (1:24:07.240)

function to be small.

Lex Fridman (1:24:08.240)

I want small loss, right?

Lex Fridman (1:24:10.880)

Well what are you going to do?

Lex Fridman (1:24:13.440)

So you sort of say, well, I'm going to average over these quantities or maximize over them

Michael I. Jordan (1:24:17.280)

or something so that, you know, I turn that uncertainty into something certain.

Lex Fridman (1:24:23.120)

So you could look at the first argument and average over it, or you could look at the

Michael I. Jordan (1:24:25.920)

second argument and average over it.

Lex Fridman (1:24:27.040)

That's Bayesian and frequentist.

Lex Fridman (1:24:28.040)

So the frequentist says, I'm going to look at the X, the data, and I'm going to take

Lex Fridman (1:24:32.840)

that as random and I'm going to average over the distribution.

Lex Fridman (1:24:35.360)

So I take the expectation loss under X. Theta is held fixed, right?

Lex Fridman (1:24:40.700)

That's called the risk.

Lex Fridman (1:24:42.140)

And so it's looking at other, all the data sets you could get, right?

Lex Fridman (1:24:46.480)

And say, how well will a certain procedure do under all those data sets?

Lex Fridman (1:24:50.200)

That's called a frequentist guarantee, right?

Lex Fridman (1:24:52.560)

So I think it is very appropriate when like you're building a piece of software and you're

Michael I. Jordan (1:24:56.080)

shipping it out there and people are using it on all kinds of data sets.

Michael I. Jordan (1:24:59.280)

You want to have a stamp, a guarantee on it that as people run it on many, many data sets

Michael I. Jordan (1:25:02.600)

that you never even thought about that 95% of the time it will do the right thing.

Lex Fridman (1:25:07.720)

Perfectly reasonable.

Michael I. Jordan (1:25:09.800)

The Bayesian perspective says, well, no, I'm going to look at the other argument of the

Lex Fridman (1:25:13.240)

loss function, the theta part, okay?

Michael I. Jordan (1:25:15.240)

That's unknown and I'm uncertain about it.

Lex Fridman (1:25:17.600)

So I could have my own personal probability for what it is, you know, how many tall people

Lex Fridman (1:25:21.560)

are there out there?

Michael I. Jordan (1:25:22.560)

I'm trying to infer the average height of the population while I have an idea roughly

Lex Fridman (1:25:25.160)

what the height is.

Lex Fridman (1:25:27.440)

So I'm going to average over the theta.

Lex Fridman (1:25:32.200)

So now that loss function as only now, again, one argument's gone, now it's a function of

Michael I. Jordan (1:25:37.760)

X and that's what a Bayesian does is they say, well, let's just focus on the particular

Michael I. Jordan (1:25:41.760)

X we got, the data set we got, we condition on that.

Lex Fridman (1:25:45.360)

Conditional on the X, I say something about my loss.

Michael I. Jordan (1:25:48.240)

That's a Bayesian approach to things.

Lex Fridman (1:25:50.480)

And the Bayesian will argue that it's not relevant to look at all the other data sets

Michael I. Jordan (1:25:54.360)

you could have gotten and average over them, the frequentist approach.

Lex Fridman (1:25:58.800)

It's really only the data sets you got, right?

Lex Fridman (1:26:02.080)

And I do agree with that, especially in situations where you're working with a scientist, you

Michael I. Jordan (1:26:06.000)

can learn a lot about the domain and you're really only focused on certain kinds of data

Lex Fridman (1:26:09.440)

and you gathered your data and you make inferences.

Michael I. Jordan (1:26:13.320)

I don't agree with it though, that, you know, in the sense that there are needs for frequentist

Michael I. Jordan (1:26:17.600)

guarantees, you're writing software, people are using it out there, you want to say something.

Lex Fridman (1:26:20.880)

So these two things have to got to fight each other a little bit, but they have to blend.

Lex Fridman (1:26:24.880)

So long story short, there's a set of ideas that are right in the middle that are called

Lex Fridman (1:26:27.880)

empirical Bayes.

Lex Fridman (1:26:29.880)

And empirical Bayes sort of starts with the Bayesian framework.

Lex Fridman (1:26:34.600)

It's kind of arguably philosophically more, you know, reasonable and kosher.

Michael I. Jordan (1:26:40.680)

Write down a bunch of the math that kind of flows from that, and then realize there's

Michael I. Jordan (1:26:44.120)

a bunch of things you don't know because it's the real world and you don't know everything.

Lex Fridman (1:26:48.040)

So you're uncertain about certain quantities.

Lex Fridman (1:26:50.160)

At that point, ask, is there a reasonable way to plug in an estimate for those things?

Michael I. Jordan (1:26:54.440)

Okay.

Lex Fridman (1:26:55.440)

And in some cases, there's quite a reasonable thing to do, to plug in, there's a natural

Michael I. Jordan (1:27:00.480)

thing you can observe in the world that you can plug in and then do a little bit more

Lex Fridman (1:27:04.000)

mathematics and assure yourself it's really good.

Lex Fridman (1:27:06.440)

So based on math or based on human expertise, what's, what, what are good?

Lex Fridman (1:27:09.800)

Oh, they're both going in.

Michael I. Jordan (1:27:10.800)

The Bayesian framework allows you to put a lot of human expertise in, but the math kind

Michael I. Jordan (1:27:16.160)

of guides you along that path and then kind of reassures you the end, you could put that

Michael I. Jordan (1:27:19.480)

stamp of approval under certain assumptions, this thing will work.

Lex Fridman (1:27:22.780)

So you asked the question, what's my favorite, you know, or what's the most surprising, nice

Michael I. Jordan (1:27:25.960)

idea.

Lex Fridman (1:27:26.960)

So one that is more accessible is something called false discovery rate, which is, you

Michael I. Jordan (1:27:31.760)

know, you're making not just one hypothesis test or making one decision, you're making

Lex Fridman (1:27:35.520)

a whole bag of them.

Lex Fridman (1:27:37.440)

And in that bag of decisions, you look at the ones where you made a discovery, you announced

Lex Fridman (1:27:41.800)

that something interesting had happened.

Michael I. Jordan (1:27:43.320)

All right.

Lex Fridman (1:27:44.320)

That's going to be some subset of your big bag.

Lex Fridman (1:27:47.160)

In the ones you made a discovery, which subset of those are bad?

Lex Fridman (1:27:50.880)

Or false, false discoveries.

Michael I. Jordan (1:27:53.320)

You'd like the fraction of your false discoveries among your discoveries to be small.

Michael I. Jordan (1:27:57.680)

That's a different criterion than accuracy or precision or recall or sensitivity and

Michael I. Jordan (1:28:02.480)

specificity.

Lex Fridman (1:28:03.480)

It's a different quantity.

Michael I. Jordan (1:28:04.960)

Those latter ones are almost all of them have more of a frequentist flavor.

Lex Fridman (1:28:09.960)

They say, given the truth is that the null hypothesis is true.

Michael I. Jordan (1:28:13.960)

Here's what accuracy I would get, or given that the alternative is true, here's what

Lex Fridman (1:28:17.400)

I would get.

Lex Fridman (1:28:18.400)

So it's kind of going forward from the state of nature to the data.

Michael I. Jordan (1:28:22.360)

The Bayesian goes the other direction from the data back to the state of nature.

Lex Fridman (1:28:25.920)

And that's actually what false discovery rate is.

Lex Fridman (1:28:28.180)

It says, given you made a discovery, okay, that's conditioned on your data.

Lex Fridman (1:28:32.680)

What's the probability of the hypothesis?

Lex Fridman (1:28:34.960)

It's going the other direction.

Lex Fridman (1:28:36.920)

And so the classical frequency look at that, well, I can't know that there's some priors

Lex Fridman (1:28:41.000)

needed in that.

Lex Fridman (1:28:42.600)

And the empirical Bayesian goes ahead and plows forward and starts writing down these formulas

Lex Fridman (1:28:47.460)

and realizes at some point, some of those things can actually be estimated in a reasonable

Michael I. Jordan (1:28:51.280)

way.

Lex Fridman (1:28:52.600)

And so it's kind of, it's a beautiful set of ideas.

Lex Fridman (1:28:54.220)

So I, this kind of line of argument has come out.

Lex Fridman (1:28:56.800)

It's not certainly mine, but it sort of came out from Robbins around 1960.

Michael I. Jordan (1:29:02.320)

Brad Efron has written beautifully about this in various papers and books.

Lex Fridman (1:29:07.320)

And the FDR is, you know, Benjamin in Israel, John Story did this Bayesian interpretation

Lex Fridman (1:29:14.120)

and so on.

Lex Fridman (1:29:15.120)

And he used to absorb these things over the years and find it a very healthy way to think

Michael I. Jordan (1:29:18.480)

about statistics.

Michael I. Jordan (1:29:21.280)

Let me ask you about intelligence to jump slightly back out into philosophy, perhaps.

Michael I. Jordan (1:29:28.240)

You said that maybe you can elaborate, but you said that defining just even the question

Lex Fridman (1:29:33.940)

of what is intelligence is a very difficult question.

Lex Fridman (1:29:38.800)

Is it a useful question?

Lex Fridman (1:29:39.800)

Do you think we'll one day understand the fundamentals of human intelligence and what

Michael I. Jordan (1:29:45.240)

it means, you know, have good benchmarks for general intelligence that we put before our

Lex Fridman (1:29:51.880)

machines?

Lex Fridman (1:29:53.520)

So I don't work on these topics so much that you're really asking the question for a psychologist

Lex Fridman (1:29:58.240)

really.

Lex Fridman (1:29:59.240)

And I studied some, but I don't consider myself at least an expert at this point.

Lex Fridman (1:30:04.440)

You know, a psychologist aims to understand human intelligence, right?

Lex Fridman (1:30:07.680)

And I think many psychologists I know are fairly humble about this.

Michael I. Jordan (1:30:10.960)

They might try to understand how a baby understands, you know, whether something's a solid or liquid

Michael I. Jordan (1:30:15.880)

or whether something's hidden or not.

Lex Fridman (1:30:18.720)

And maybe how a child starts to learn the meaning of certain words, what's a verb, what's

Michael I. Jordan (1:30:24.400)

a noun and also, you know, slowly but surely trying to figure out things.

Lex Fridman (1:30:30.580)

But humans ability to take a really complicated environment, reason about it, abstract about

Michael I. Jordan (1:30:35.720)

it, find the right abstractions, communicate about it, interact and so on is just, you

Lex Fridman (1:30:41.520)

know, really staggeringly rich and complicated.

Lex Fridman (1:30:46.920)

And so, you know, I think in all humility, we don't think we're kind of aiming for that

Lex Fridman (1:30:51.320)

in the near future.

Michael I. Jordan (1:30:52.320)

A certain psychologist doing experiments with babies in the lab or with people talking has

Lex Fridman (1:30:56.820)

a much more limited aspiration.

Lex Fridman (1:30:58.920)

And you know, Kahneman and Tversky would look at our reasoning patterns and they're not

Michael I. Jordan (1:31:02.120)

deeply understanding all the how we do our reasoning, but they're sort of saying, hey,

Michael I. Jordan (1:31:05.880)

here's some oddities about the reasoning and some things you should think about it.

Lex Fridman (1:31:09.480)

But also, as I emphasize in some things I've been writing about, you know, AI, the revolution

Michael I. Jordan (1:31:14.560)

hasn't happened yet.

Lex Fridman (1:31:15.560)

Yeah.

Michael I. Jordan (1:31:16.560)

Great blog post.

Michael I. Jordan (1:31:17.560)

I've been emphasizing that, you know, if you step back and look at intelligent systems

Michael I. Jordan (1:31:22.580)

of any kind and whatever you mean by intelligence, it's not just the humans or the animals or,

Michael I. Jordan (1:31:26.800)

you know, the plants or whatever, you know, so a market that brings goods into a city,

Michael I. Jordan (1:31:31.680)

you know, food to restaurants or something every day is a system.

Lex Fridman (1:31:35.680)

It's a decentralized set of decisions.

Michael I. Jordan (1:31:37.820)

Looking at it from far enough away, it's just like a collection of neurons.

Lex Fridman (1:31:40.840)

Every neuron is making its own little decisions, presumably in some way.

Lex Fridman (1:31:44.600)

And if you step back enough, every little part of an economic system is making all of

Lex Fridman (1:31:48.000)

its decisions.

Lex Fridman (1:31:49.560)

And just like with the brain, who knows what an individual neuron does and what the overall

Lex Fridman (1:31:53.020)

goal is, right?

Lex Fridman (1:31:54.800)

But something happens at some aggregate level, same thing with the economy.

Lex Fridman (1:31:58.560)

People eat in a city and it's robust.

Michael I. Jordan (1:32:01.380)

It works at all scales, small villages to big cities.

Lex Fridman (1:32:04.840)

It's been working for thousands of years.

Michael I. Jordan (1:32:07.040)

It works rain or shine, so it's adaptive.

Lex Fridman (1:32:10.520)

So all the kind of, you know, those are adjectives one tends to apply to intelligent systems.

Michael I. Jordan (1:32:14.680)

Robust, adaptive, you know, you don't need to keep adjusting it, self healing, whatever.

Lex Fridman (1:32:19.960)

Plus not perfect.

Michael I. Jordan (1:32:20.960)

You know, intelligences are never perfect and markets are not perfect.

Lex Fridman (1:32:24.680)

But I do not believe in this era that you cannot, that you can say, well, our computers

Michael I. Jordan (1:32:28.160)

are, our humans are smart, but you know, no markets are not, more markets are.

Lex Fridman (1:32:31.760)

So they are intelligent.

Michael I. Jordan (1:32:34.080)

Now we humans didn't evolve to be markets.

Lex Fridman (1:32:38.160)

We've been participating in them, right?

Lex Fridman (1:32:40.320)

But we are not ourselves a market per se.

Lex Fridman (1:32:43.280)

The neurons could be viewed as the market.

Michael I. Jordan (1:32:45.920)

There's economic, you know, neuroscience kind of perspective.

Lex Fridman (1:32:48.200)

That's interesting to pursue all that.

Michael I. Jordan (1:32:50.320)

The point though is, is that if you were to study humans and really be the world's best

Michael I. Jordan (1:32:54.200)

psychologist studied for thousands of years and come up with the theory of human intelligence,

Michael I. Jordan (1:32:57.440)

you might have never discovered principles of markets, you know, supply demand curves

Lex Fridman (1:33:01.840)

and you know, matching and auctions and all that.

Michael I. Jordan (1:33:05.000)

Those are real principles and they lead to a form of intelligence that's not maybe human

Lex Fridman (1:33:08.760)

intelligence.

Michael I. Jordan (1:33:09.760)

It's arguably another kind of intelligence.

Michael I. Jordan (1:33:11.480)

There probably are third kinds of intelligence or fourth that none of us are really thinking

Michael I. Jordan (1:33:14.880)

too much about right now.

Lex Fridman (1:33:16.480)

So if you really, and then all of those are relevant to computer systems in the future.

Michael I. Jordan (1:33:20.840)

Certainly the market one is relevant right now.

Michael I. Jordan (1:33:23.880)

Whereas the understanding of human intelligence is not so clear that it's relevant right now.

Michael I. Jordan (1:33:27.440)

Probably not.

Lex Fridman (1:33:29.360)

So if you want general intelligence, whatever one means by that, or, you know, understanding

Michael I. Jordan (1:33:33.160)

intelligence in a deep sense and all that, it is definitely has to be not just human

Lex Fridman (1:33:37.000)

intelligence.

Michael I. Jordan (1:33:38.000)

It's gotta be this broader thing.

Lex Fridman (1:33:39.280)

And that's not a mystery.

Michael I. Jordan (1:33:40.480)

Markets are intelligent.

Michael I. Jordan (1:33:41.480)

So, you know, it's definitely not just a philosophical stance to say we've got to move beyond intelligence.

Michael I. Jordan (1:33:46.000)

That sounds ridiculous.

Lex Fridman (1:33:47.000)

Yeah.

Lex Fridman (1:33:48.000)

But it's not.

Lex Fridman (1:33:49.000)

And in that blog post, you define different kinds of like intelligent infrastructure,

Michael I. Jordan (1:33:52.160)

AI, which I really like is some of the concepts you've just been describing.

Lex Fridman (1:33:58.040)

Do you see ourselves, if we see earth, human civilization as a single organism, do you

Michael I. Jordan (1:34:02.720)

think the intelligence of that organism, when you think from the perspective of markets

Lex Fridman (1:34:06.980)

and intelligence infrastructure is increasing, is it increasing linearly?

Lex Fridman (1:34:12.340)

Is it increasing exponentially?

Lex Fridman (1:34:14.240)

What do you think the future of that intelligence?

Michael I. Jordan (1:34:16.000)

Yeah, I don't know.

Michael I. Jordan (1:34:17.000)

I don't tend to think, I don't tend to answer questions like that because you know, that's

Michael I. Jordan (1:34:20.560)

science fiction.

Lex Fridman (1:34:21.560)

I'm hoping to catch you off guard.

Michael I. Jordan (1:34:25.200)

Well again, because you said it's so far in the future, it's fun to ask and you'll probably,

Lex Fridman (1:34:31.320)

you know, like you said, predicting the future is really nearly impossible.

Lex Fridman (1:34:36.440)

But say as an axiom, one day we create a human level, a superhuman level intelligent, not

Lex Fridman (1:34:43.720)

the scale of markets, but the scale of an individual.

Lex Fridman (1:34:47.560)

What do you think it is, what do you think it would take to do that?

Michael I. Jordan (1:34:51.760)

Or maybe to ask another question is how would that system be different than the biological

Lex Fridman (1:34:58.880)

human beings that we see around us today?

Lex Fridman (1:35:01.480)

Is it possible to say anything interesting to that question or is it just a stupid question?

Michael I. Jordan (1:35:06.160)

It's not a stupid question, but it's science fiction.

Lex Fridman (1:35:08.200)

Science fiction.

Lex Fridman (1:35:09.200)

And so I'm totally happy to read science fiction and think about it from time in my own life.

Michael I. Jordan (1:35:13.400)

I loved, there was this like brain in a vat kind of, you know, little thing that people

Michael I. Jordan (1:35:17.480)

were talking about when I was a student, I remember, you know, imagine that, you know,

Lex Fridman (1:35:22.680)

between your brain and your body, there's a, you know, there's a bunch of wires, right?

Lex Fridman (1:35:26.960)

And suppose that every one of them was replaced with a literal wire.

Lex Fridman (1:35:31.480)

And then suppose that wire was turned in actually a little wireless, you know, there's a receiver

Lex Fridman (1:35:35.000)

and sender.

Lex Fridman (1:35:36.000)

So the brain has got all the senders and receiver, you know, on all of its exiting, you know,

Michael I. Jordan (1:35:41.560)

axons and all the dendrites down to the body have replaced with senders and receivers.

Lex Fridman (1:35:45.920)

Now you could move the body off somewhere and put the brain in a vat, right?

Lex Fridman (1:35:50.080)

And then you could do things like start killing off those senders and receivers one by one.

Lex Fridman (1:35:54.600)

And after you've killed off all of them, where is that person?

Michael I. Jordan (1:35:56.960)

You know, they thought they were out in the body walking around the world and they moved

Lex Fridman (1:35:59.640)

on.

Lex Fridman (1:36:00.640)

So those are science fiction things.

Lex Fridman (1:36:01.640)

Those are fun to think about.

Michael I. Jordan (1:36:02.640)

It's just intriguing about where is, what is thought, where is it and all that.

Lex Fridman (1:36:05.760)

And I think every 18 year old should take philosophy classes and think about these things.

Lex Fridman (1:36:10.680)

And I think that everyone should think about what could happen in society that's kind of

Lex Fridman (1:36:13.440)

bad and all that.

Lex Fridman (1:36:14.440)

But I really don't think that's the right thing for most of us that are my age group

Lex Fridman (1:36:17.600)

to be doing and thinking about.

Michael I. Jordan (1:36:19.480)

I really think that we have so many more present, you know, first challenges and dangers and

Michael I. Jordan (1:36:26.720)

real things to build and all that such that, you know, spending too much time on science

Michael I. Jordan (1:36:32.320)

fiction, at least in public for like this, I think is not what we should be doing.

Lex Fridman (1:36:36.080)

Maybe over beers in private.

Michael I. Jordan (1:36:37.600)

That's right.

Michael I. Jordan (1:36:38.600)

Well, I'm not going to broadcast where I have beers because this is going to go on Facebook

Lex Fridman (1:36:43.600)

and I don't want a lot of people showing up there.

Lex Fridman (1:36:45.480)

But yeah, I'll, I love Facebook, Twitter, Amazon, YouTube.

Michael I. Jordan (1:36:51.640)

I have I'm optimistic and hopeful, but maybe, maybe I don't have grounds for such optimism

Lex Fridman (1:36:58.280)

and hope.

Lex Fridman (1:36:59.280)

But let me ask, you've mentored some of the brightest sort of some of the seminal figures

Lex Fridman (1:37:07.160)

in the field.

Lex Fridman (1:37:08.160)

Can you give advice to people who are undergraduates today?

Lex Fridman (1:37:14.080)

What does it take to take, you know, advice on their journey if they're interested in

Michael I. Jordan (1:37:17.640)

machine learning and in the ideas of markets from economics and psychology and all the

Lex Fridman (1:37:23.920)

kinds of things that you've exploring?

Lex Fridman (1:37:25.680)

What steps should they take on that journey?

Lex Fridman (1:37:27.960)

Well, yeah, first of all, the door is open and second, it's a journey.

Michael I. Jordan (1:37:30.360)

I like your language there.

Michael I. Jordan (1:37:33.880)

It is not that you're so brilliant and you have great, brilliant ideas and therefore

Michael I. Jordan (1:37:37.120)

that's just, you know, that's how you have success or that's how you enter into the field.

Michael I. Jordan (1:37:42.440)

It's that you apprentice yourself, you spend a lot of time, you work on hard things, you

Michael I. Jordan (1:37:48.480)

try and pull back and you be as broad as you can, you talk to lots of people.

Lex Fridman (1:37:53.880)

And it's like entering in any kind of a creative community.

Michael I. Jordan (1:37:57.000)

There's years that are needed and human connections are critical to it.

Michael I. Jordan (1:38:01.600)

So, you know, I think about, you know, being a musician or being an artist or something,

Michael I. Jordan (1:38:06.080)

you don't just, you know, immediately from day one, you know, you're a genius and therefore

Lex Fridman (1:38:10.600)

you do it.

Michael I. Jordan (1:38:11.600)

No, you, you know, practice really, really hard on basics and you be humble about where

Lex Fridman (1:38:18.900)

you are and then, and you realize you'll never be an expert on everything.

Lex Fridman (1:38:22.200)

So you kind of pick and there's a lot of randomness and a lot of kind of luck, but luck just kind

Michael I. Jordan (1:38:29.460)

of picks out which branch of the tree you go down, but you'll go down some branch.

Lex Fridman (1:38:33.960)

So yeah, it's a community.

Lex Fridman (1:38:35.460)

So the graduate school is, I still think is one of the wonderful phenomena that we have

Michael I. Jordan (1:38:39.200)

in our, in our world.

Lex Fridman (1:38:40.780)

It's very much about apprenticeship with an advisor.

Michael I. Jordan (1:38:43.160)

It's very much about a group of people you belong to.

Lex Fridman (1:38:45.780)

It's a four or five year process.

Lex Fridman (1:38:47.020)

So it's plenty of time to start from kind of nothing to come up to something, you know,

Michael I. Jordan (1:38:51.580)

more, more expertise, and then to start to have your own creativity start to flower,

Michael I. Jordan (1:38:54.700)

even surprising your own self.

Lex Fridman (1:38:58.240)

And it's a very cooperative endeavor.

Michael I. Jordan (1:38:59.760)

I think a lot of people think of science as highly competitive and I think in some other

Lex Fridman (1:39:05.620)

fields it might be more so.

Michael I. Jordan (1:39:08.080)

Here it's way more cooperative than you might imagine.

Lex Fridman (1:39:11.860)

And people are always teaching each other something and people are always more than

Michael I. Jordan (1:39:14.660)

happy to be clear that, so I feel I'm an expert on certain kinds of things, but I'm very much

Michael I. Jordan (1:39:20.000)

not expert on lots of other things and a lot of them are relevant and a lot of them are,

Michael I. Jordan (1:39:23.480)

I should know, but should in some society, you know, you don't.

Lex Fridman (1:39:26.320)

So I'm always willing to reveal my ignorance to people around me so they can teach me things.

Lex Fridman (1:39:32.100)

And I think a lot of us feel that way about our field.

Lex Fridman (1:39:34.220)

So it's very cooperative.

Michael I. Jordan (1:39:35.460)

I might add it's also very international because it's so cooperative.

Lex Fridman (1:39:39.140)

We see no barriers.

Lex Fridman (1:39:40.780)

And so that the nationalism that you see, especially in the current era and everything

Michael I. Jordan (1:39:44.460)

is just at odds with the way that most of us think about what we're doing here, where

Michael I. Jordan (1:39:48.180)

this is a human endeavor and we cooperate and are very much trying to do it together

Lex Fridman (1:39:53.420)

for the, you know, the benefit of everybody.

Lex Fridman (1:39:56.580)

So last question, where and how and why did you learn French and which language is more

Lex Fridman (1:40:02.820)

beautiful English or French?

Michael I. Jordan (1:40:05.660)

Great question.

Lex Fridman (1:40:06.660)

So first of all, I think Italian is actually more beautiful than French and English.

Lex Fridman (1:40:10.100)

And I also speak that.

Lex Fridman (1:40:11.100)

So I'm married to an Italian and I have kids and we speak Italian.

Michael I. Jordan (1:40:15.860)

Anyway, all kidding aside, every language allows you to express things a bit differently.

Lex Fridman (1:40:23.180)

And it is one of the great fun things to do in life is to explore those things.

Lex Fridman (1:40:26.820)

So in fact, when I kids or teens or college students ask me what they study, I say, well,

Lex Fridman (1:40:34.540)

do what your heart, where your heart is, certainly do a lot of math.

Michael I. Jordan (1:40:36.980)

Math is good for everybody, but do some poetry and do some history and do some language too.

Lex Fridman (1:40:42.500)

You know, throughout your life, you'll want to be a thinking person.

Michael I. Jordan (1:40:44.620)

You'll want to have done that.

Michael I. Jordan (1:40:47.500)

For me, French I learned when I was, I'd say a late teen, I was living in the middle of

Michael I. Jordan (1:40:54.700)

the country in Kansas and not much was going on in Kansas with all due respect to Kansas.

Lex Fridman (1:41:01.100)

And so my parents happened to have some French books on the shelf and just in my boredom,

Michael I. Jordan (1:41:04.380)

I pulled them down and I found this is fun.

Lex Fridman (1:41:07.140)

And I kind of learned the language by reading.

Lex Fridman (1:41:09.220)

And when I first heard it spoken, I had no idea what was being spoken, but I realized

Lex Fridman (1:41:13.540)

I had somehow knew it from some previous life and so I made the connection.

Lex Fridman (1:41:18.540)

But then I traveled and just I love to go beyond my own barriers and my own comfort

Lex Fridman (1:41:23.500)

or whatever.

Lex Fridman (1:41:24.500)

And I found myself on trains in France next to say older people who had lived a whole

Lex Fridman (1:41:29.460)

life of their own.

Lex Fridman (1:41:30.460)

And the ability to communicate with them was special and the ability to also see myself

Michael I. Jordan (1:41:37.900)

in other people's shoes and have empathy and kind of work on that language as part of that.

Lex Fridman (1:41:43.100)

So after that kind of experience and also embedding myself in French culture, which

Michael I. Jordan (1:41:49.140)

is quite amazing, languages are rich, not just because there's something inherently

Michael I. Jordan (1:41:53.780)

beautiful about it, but it's all the creativity that went into it.

Lex Fridman (1:41:55.940)

So I learned a lot of songs, read poems, read books.

Lex Fridman (1:41:59.900)

And then I was here actually at MIT where we're doing the podcast today and a young

Lex Fridman (1:42:05.300)

professor not yet married and not having a lot of friends in the area.

Lex Fridman (1:42:11.960)

So I just didn't have, I was kind of a bored person.

Lex Fridman (1:42:13.980)

I said, I heard a lot of Italians around.

Michael I. Jordan (1:42:16.020)

There's happened to be a lot of Italians at MIT, an Italian professor for some reason.

Lex Fridman (1:42:20.020)

And so I was kind of vaguely understanding what they were talking about.

Michael I. Jordan (1:42:22.060)

I said, well, I should learn this language too.

Lex Fridman (1:42:23.660)

So I did.

Lex Fridman (1:42:25.620)

And then later met my spouse and Italian became a part of my life.

Lex Fridman (1:42:30.860)

But I go to China a lot these days.

Michael I. Jordan (1:42:32.180)

I go to Asia, I go to Europe and every time I go, I kind of am amazed by the richness

Michael I. Jordan (1:42:38.160)

of human experience and the people don't have any idea if you haven't traveled, kind of

Lex Fridman (1:42:42.820)

how amazingly rich and I love the diversity.

Lex Fridman (1:42:46.900)

It's not just a buzzword to me.

Michael I. Jordan (1:42:48.060)

It really means something.

Lex Fridman (1:42:49.060)

I love to embed myself with other people's experiences.

Lex Fridman (1:42:53.180)

And so yeah, learning language is a big part of that.

Michael I. Jordan (1:42:56.420)

I think I've said in some interview at some point that if I had millions of dollars and

Michael I. Jordan (1:43:00.460)

infinite time or whatever, what would you really work on if you really wanted to do

Lex Fridman (1:43:03.300)

AI?

Lex Fridman (1:43:04.300)

And for me, that is natural language and really done right.

Lex Fridman (1:43:07.360)

Deep understanding of language.

Michael I. Jordan (1:43:09.840)

That's to me, an amazingly interesting scientific challenge.

Lex Fridman (1:43:13.580)

One we're very far away on.

Michael I. Jordan (1:43:15.180)

One we're very far away, but good natural language.

Lex Fridman (1:43:17.720)

People are kind of really invested then.

Michael I. Jordan (1:43:19.140)

I think a lot of them see that's where the core of AI is that if you understand that

Michael I. Jordan (1:43:22.460)

you really help human communication, you understand something about the human mind, the semantics

Michael I. Jordan (1:43:26.580)

that come out of the human mind and I agree, I think that will be such a long time.

Lex Fridman (1:43:30.980)

So I didn't do that in my career just cause I kind of, I was behind in the early days.

Michael I. Jordan (1:43:34.720)

I didn't kind of know enough of that stuff.

Michael I. Jordan (1:43:36.460)

I was at MIT, I didn't learn much language and it was too late at some point to kind

Michael I. Jordan (1:43:41.180)

of spend a whole career doing that, but I admire that field and so in my little way

Michael I. Jordan (1:43:47.180)

by learning language, you know, kind of that part of my brain has been trained up.

Michael I. Jordan (1:43:53.340)

Jan was right.

Lex Fridman (1:43:55.460)

You truly are the Miles Davis of machine learning.

Michael I. Jordan (1:43:57.460)

I don't think there's a better place than it.

Lex Fridman (1:43:59.620)

Mike it was a huge honor talking to you today.

Michael I. Jordan (1:44:01.580)

Merci beaucoup.

Lex Fridman (1:44:02.580)

All right.

Michael I. Jordan (1:44:03.580)

It's been my pleasure.

Michael I. Jordan (1:44:04.580)

Thanks for listening to this conversation with Michael I. Jordan and thank you to our

Michael I. Jordan (1:44:09.300)

presenting sponsor, Cash App.

Michael I. Jordan (1:44:11.420)

Download it, use code LEXPodcast, you'll get $10 and $10 will go to FIRST, an organization

Michael I. Jordan (1:44:18.100)

that inspires and educates young minds to become science and technology innovators of

Lex Fridman (1:44:22.580)

tomorrow.

Michael I. Jordan (1:44:23.880)

If you enjoy this podcast, subscribe on YouTube, give it five stars on Apple Podcast, support

Lex Fridman (1:44:28.820)

on Patreon, or simply connect with me on Twitter at Lex Friedman.

Lex Fridman (1:44:34.700)

And now let me leave you with some words of wisdom from Michael I. Jordan from his blog

Michael I. Jordan (1:44:39.340)

post titled Artificial Intelligence, the revolution hasn't happened yet, calling for broadening

Michael I. Jordan (1:44:45.580)

the scope of the AI field.

Michael I. Jordan (1:44:48.560)

We should embrace the fact that what we are witnessing is the creation of a new branch

Michael I. Jordan (1:44:52.860)

of engineering.

Michael I. Jordan (1:44:54.340)

The term engineering is often invoked in a narrow sense in academia and beyond with overtones

Michael I. Jordan (1:45:00.660)

of cold, effectless machinery and negative connotations of loss of control by humans.

Lex Fridman (1:45:07.540)

But an engineering discipline can be what we want it to be.

Michael I. Jordan (1:45:11.020)

In the current era, we have a real opportunity to conceive of something historically new,

Lex Fridman (1:45:16.340)

a human centric engineering discipline.

Michael I. Jordan (1:45:19.740)

I will resist giving this emerging discipline a name, but if the acronym AI continues to

Lex Fridman (1:45:24.380)

be used, let's be aware of the very real limitations of this placeholder.

Michael I. Jordan (1:45:29.860)

Let's broaden our scope, tone down the hype, and recognize the serious challenges ahead.

Lex Fridman (1:45:37.300)

Thank you for listening and hope to see you next time.

Lex Fridman (20:02.160)

you say is the most interesting disagreement you have with Jan Lacune?

Lex Fridman (20:07.400)

So Jan's an old friend and I just say that I don't think we disagree about very much

Michael I. Jordan (20:12.520)

really.

Michael I. Jordan (20:13.520)

He and I both kind of have a let's build it kind of mentality and does it work kind of

Michael I. Jordan (20:18.800)

mentality and kind of concrete.

Michael I. Jordan (20:21.360)

We both speak French and we speak French more together and we have a lot in common.

Lex Fridman (20:27.120)

And so if one wanted to highlight a disagreement, it's not really a fundamental one.

Lex Fridman (20:31.800)

I think it's just kind of what we're emphasizing.

Michael I. Jordan (20:35.200)

Jan has emphasized pattern recognition and has emphasized prediction.

Lex Fridman (20:43.440)

And it's interesting to try to take that as far as you can.

Lex Fridman (20:45.320)

If you could do perfect prediction, what would that give you kind of as a thought experiment?

Lex Fridman (20:50.600)

And I think that's way too limited.

Michael I. Jordan (20:55.200)

We cannot do perfect prediction.

Michael I. Jordan (20:56.640)

We will never have the data sets that allow me to figure out what you're about ready to

Michael I. Jordan (20:59.360)

do, what question you're going to ask next.

Lex Fridman (21:00.760)

I have no clue.

Michael I. Jordan (21:01.760)

I will never know such things.

Michael I. Jordan (21:03.320)

Moreover, most of us find ourselves during the day in all kinds of situations we had

Michael I. Jordan (21:07.520)

no anticipation of that are kind of very, very novel in various ways.

Lex Fridman (21:13.480)

And in that moment, we want to think through what we want.

Lex Fridman (21:16.340)

And also there's going to be market forces acting on us.

Michael I. Jordan (21:19.240)

I'd like to go down that street, but now it's full because there's a crane in the street.

Michael I. Jordan (21:22.320)

I got it.

Lex Fridman (21:23.320)

I got to think about that.

Michael I. Jordan (21:24.320)

I got to think about what I might really want here.

Lex Fridman (21:26.240)

And I got to sort of think about how much it costs me to do this action versus this

Michael I. Jordan (21:29.520)

action.

Lex Fridman (21:30.520)

I got to think about the risks involved.

Michael I. Jordan (21:32.800)

A lot of our current pattern recognition and prediction systems don't do any risk evaluations.

Lex Fridman (21:37.000)

They have no error bars, right?

Michael I. Jordan (21:39.000)

I got to think about other people's decisions around me.

Michael I. Jordan (21:41.080)

I got to think about a collection of my decisions, even just thinking about like a medical treatment,

Michael I. Jordan (21:45.560)

you know, I'm not going to take a, the prediction of a neural net about my health, about something

Lex Fridman (21:50.480)

consequential.

Michael I. Jordan (21:51.480)

I'm not about ready to have a heart attack because some number is over 0.7.

Michael I. Jordan (21:54.580)

Even if you had all the data in the world that ever been collected about heart attacks

Michael I. Jordan (21:58.920)

better than any doctor ever had, I'm not going to trust the output of that neural net to

Lex Fridman (22:02.640)

predict my heart attack.

Michael I. Jordan (22:03.640)

I'm going to want to ask what if questions around that.

Michael I. Jordan (22:06.400)

I'm going to want to look at some us or other possible data I didn't have, causal things.

Michael I. Jordan (22:10.360)

I'm going to want to have a dialogue with a doctor about things we didn't think about

Lex Fridman (22:13.680)

when he gathered the data.

Michael I. Jordan (22:15.480)

You know, I could go on and on.

Lex Fridman (22:16.640)

I hope you can see.

Lex Fridman (22:17.900)

And I don't, I think that if you say predictions, everything that, that, that you're missing

Lex Fridman (22:21.520)

all of this stuff.

Lex Fridman (22:23.520)

And so prediction plus decision making is everything, but both of them are equally important.

Lex Fridman (22:28.240)

And so the field has emphasized prediction, Jan rightly so has seen how powerful that

Michael I. Jordan (22:32.520)

is.

Lex Fridman (22:33.660)

But at the cost of people not being aware that decision making is where the rubber really

Michael I. Jordan (22:37.240)

hits the road, where human lives are at stake, where risks are being taken, where you got

Lex Fridman (22:41.440)

to gather more data.

Michael I. Jordan (22:42.440)

You got to think about the error bars.

Lex Fridman (22:43.640)

You got to think about the consequences of your decisions on others.

Michael I. Jordan (22:45.920)

You got to think about the economy around your decisions, blah, blah, blah, blah.

Lex Fridman (22:48.960)

I'm not the only one working on those, but we're a smaller tribe.

Lex Fridman (22:52.120)

And right now we're not the one that people talk about the most.

Lex Fridman (22:56.400)

But you know, if you go out in the real world and industry, you know, at Amazon, I'd say

Michael I. Jordan (23:00.460)

half the people there are working on decision making and the other half are doing, you know,

Lex Fridman (23:03.720)

the pattern recognition.

Michael I. Jordan (23:04.720)

It's important.

Lex Fridman (23:05.720)

And the words of pattern recognition and prediction, I think the distinction there, not to linger

Michael I. Jordan (23:10.160)

on words, but the distinction there is more a constrained sort of in the lab data set

Michael I. Jordan (23:16.120)

versus decision making is talking about consequential decisions in the real world, under the messiness

Lex Fridman (23:21.160)

and the uncertainty of the real world.

Lex Fridman (23:23.760)

And just the whole of it, the whole mess of it that actually touches human beings and

Michael I. Jordan (23:27.480)

scale.

Lex Fridman (23:28.480)

And the forces, that's the distinction.

Michael I. Jordan (23:31.120)

It helps add those, that perspective, that broader perspective.

Lex Fridman (23:33.840)

You're right.

Michael I. Jordan (23:34.840)

I totally agree.

Michael I. Jordan (23:35.840)

On the other hand, if you're a real prediction person, of course, you want it to be in the

Michael I. Jordan (23:38.120)

real world.

Lex Fridman (23:39.120)

You want to predict real world events.

Michael I. Jordan (23:40.120)

I'm just saying that's not possible with just data sets.

Michael I. Jordan (23:43.200)

That it has to be in the context of, you know, strategic things that someone's doing, data

Michael I. Jordan (23:47.520)

they might gather, things they could have gathered, the reasoning process around data.

Lex Fridman (23:50.880)

It's not just taking data and making predictions based on the data.

Lex Fridman (23:53.580)

So one of the things that you're working on, I'm sure there's others working on it, but

Michael I. Jordan (23:58.280)

I don't hear often it talked about, especially in the clarity that you talk about it, and

Michael I. Jordan (24:04.960)

I think it's both the most exciting and the most concerning area of AI in terms of decision

Lex Fridman (24:10.600)

making.

Lex Fridman (24:11.600)

So you've talked about AI systems that help make decisions that scale in a distributed

Lex Fridman (24:15.400)

way, millions, billions decisions, sort of markets of decisions.

Lex Fridman (24:19.720)

Can you, as a starting point, sort of give an example of a system that you think about

Lex Fridman (24:24.920)

when you're thinking about these kinds of systems?

Michael I. Jordan (24:27.240)

Yeah, so first of all, you're absolutely getting into some territory, which I will be beyond

Lex Fridman (24:31.400)

my expertise.

Lex Fridman (24:32.400)

And there are lots of things that are going to be very not obvious to think about.

Michael I. Jordan (24:35.720)

Just like, again, I like to think about history a little bit, but think about put yourself

Michael I. Jordan (24:39.920)

back in the sixties.

Lex Fridman (24:40.920)

There was kind of a banking system that wasn't computerized really.

Michael I. Jordan (24:43.440)

There was database theory emerging and database people had to think about how do I actually

Michael I. Jordan (24:48.160)

not just move data around, but actual money and have it be, you know, valid and have transactions

Michael I. Jordan (24:53.560)

that ATMs happen that are actually, you know, all valid and so on and so forth.

Lex Fridman (24:57.840)

So that's the kind of issues you get into when you start to get serious about sorts

Michael I. Jordan (25:01.560)

of things like this.

Michael I. Jordan (25:02.960)

I like to think about as kind of almost a thought experiment to help me think something

Michael I. Jordan (25:07.240)

simpler, which is the music market.

Lex Fridman (25:11.160)

And because there is, to first order, there is no music market in the world right now

Lex Fridman (25:16.160)

and in our country, for sure.

Michael I. Jordan (25:18.740)

There are something called things called record companies and they make money and they prop

Michael I. Jordan (25:23.720)

up a few really good musicians and make them superstars and they all make huge amounts

Lex Fridman (25:29.480)

of money.

Lex Fridman (25:30.980)

But there's a long tail of huge numbers of people that make lots and lots of really good

Lex Fridman (25:33.820)

music that is actually listened to by more people than the famous people.

Michael I. Jordan (25:40.560)

They are not in a market.

Lex Fridman (25:41.560)

They cannot have a career.

Michael I. Jordan (25:42.820)

They do not make money.

Michael I. Jordan (25:43.880)

The creators, the creators, the creators, the so called influencers or whatever that

Michael I. Jordan (25:47.760)

diminishes who they are.

Lex Fridman (25:49.340)

So there are people who make extremely good music, especially in the hip hop or Latin

Michael I. Jordan (25:53.360)

world these days.

Lex Fridman (25:55.260)

They do it on their laptop.

Michael I. Jordan (25:56.320)

That's what they do on the weekend and they have another job during the week and they

Lex Fridman (26:01.040)

put it up on SoundCloud or other sites.

Michael I. Jordan (26:03.920)

Eventually it gets streamed.

Lex Fridman (26:04.920)

It now gets turned into bits.

Michael I. Jordan (26:06.140)

It's not economically valuable.

Lex Fridman (26:07.720)

The information is lost.

Michael I. Jordan (26:08.980)

It gets put up there.

Lex Fridman (26:10.200)

People stream it.

Michael I. Jordan (26:11.580)

You walk around in a big city, you see people with headphones, especially young kids listening

Lex Fridman (26:16.240)

to music all the time.

Michael I. Jordan (26:17.240)

If you look at the data, very little of the music they are listening to is the famous

Lex Fridman (26:21.080)

people's music and none of it's old music.

Michael I. Jordan (26:23.120)

It's all the latest stuff.

Lex Fridman (26:24.360)

But the people who made that latest stuff are like some 16 year old somewhere who will

Michael I. Jordan (26:27.480)

never make a career out of this, who will never make money.

Lex Fridman (26:29.600)

Of course there will be a few counter examples.

Michael I. Jordan (26:31.480)

The record companies incentivize to pick out a few and highlight them.

Lex Fridman (26:35.360)

Long story short, there's a missing market there.

Michael I. Jordan (26:37.720)

There is not a consumer producer relationship at the level of the actual creative acts.

Michael I. Jordan (26:43.480)

The pipelines and Spotify's of the world that take this stuff and stream it along, they

Michael I. Jordan (26:48.200)

make money off of subscriptions or advertising and those things.

Lex Fridman (26:51.160)

They're making the money.

Michael I. Jordan (26:52.160)

All right.

Lex Fridman (26:53.160)

And then they will offer bits and pieces of it to a few people again to highlight that

Michael I. Jordan (26:55.800)

they simulate a market.

Michael I. Jordan (26:58.640)

Anyway, a real market would be if you're a creator of music that you actually are somebody

Michael I. Jordan (27:03.560)

who's good enough that people want to listen to you, you should have the data available

Lex Fridman (27:07.440)

to you.

Michael I. Jordan (27:08.440)

There should be a dashboard showing a map of the United States.

Lex Fridman (27:11.480)

So in last week, here's all the places your songs were listened to.

Michael I. Jordan (27:14.680)

It should be transparent, vetable, so that if someone down in Providence sees that you're

Lex Fridman (27:20.520)

being listened to 10,000 times in Providence, that they know that's real data.

Michael I. Jordan (27:24.160)

You know it's real data.

Lex Fridman (27:25.320)

They will have you come give a show down there.

Michael I. Jordan (27:27.300)

They will broadcast to the people who've been listening to you that you're coming.

Lex Fridman (27:30.040)

If you do this right, you could go down there and make $20,000.

Michael I. Jordan (27:34.480)

You do that three times a year, you start to have a career.

Lex Fridman (27:37.100)

So in this sense, AI creates jobs.

Michael I. Jordan (27:39.600)

It's not about taking away human jobs.

Lex Fridman (27:40.680)

It's creating new jobs because it creates a new market.

Michael I. Jordan (27:43.480)

Once you've created a market, you've now connected up producers and consumers.

Michael I. Jordan (27:46.800)

The person who's making the music can say to someone who comes to their shows a lot,

Michael I. Jordan (27:50.000)

hey, I'll play at your daughter's wedding for $10,000.

Lex Fridman (27:53.200)

You'll say 8,000.

Michael I. Jordan (27:54.200)

They'll say 9,000.

Lex Fridman (27:55.200)

Then again, you can now get an income up to $100,000.

Michael I. Jordan (27:59.000)

You're not going to be a millionaire.

Lex Fridman (28:01.920)

And now even think about really the value of music is in these personal connections,

Michael I. Jordan (28:06.900)

even so much so that a young kid wants to wear a tshirt with their favorite musician's

Lex Fridman (28:13.180)

signature on it.

Lex Fridman (28:14.840)

So if they listen to the music on the internet, the internet should be able to provide them

Lex Fridman (28:18.080)

with a button that they push and the merchandise arrives the next day.

Michael I. Jordan (28:21.840)

We can do that.

Lex Fridman (28:23.000)

And now why should we do that?

Michael I. Jordan (28:24.400)

Well, because the kid who bought the shirt will be happy, but more the person who made

Lex Fridman (28:27.560)

the music will get the money.

Michael I. Jordan (28:29.080)

There's no advertising needed.

Lex Fridman (28:32.360)

So you can create markets between producers and consumers, take 5% cut.

Michael I. Jordan (28:36.460)

Your company will be perfectly sound.

Michael I. Jordan (28:39.200)

It'll go forward into the future and it will create new markets and that raises human happiness.

Michael I. Jordan (28:45.080)

Now this seems like, well, this is easy, just create this dashboard, kind of create some

Lex Fridman (28:48.280)

connections and all that.

Lex Fridman (28:49.280)

But if you think about Uber or whatever, you think about the challenges in the real world

Michael I. Jordan (28:52.900)

of doing things like this, and there are actually new principles going to be needed.

Michael I. Jordan (28:56.180)

You're trying to create a new kind of two way market at a different scale that's ever

Lex Fridman (28:59.080)

been done before.

Michael I. Jordan (29:00.080)

There's going to be unwanted aspects of the market.

Lex Fridman (29:04.720)

There'll be bad people.

Michael I. Jordan (29:05.720)

There'll be the data will get used in the wrong ways, it'll fail in some ways, it won't

Lex Fridman (29:10.880)

deliver about.

Michael I. Jordan (29:11.880)

You have to think that through.

Michael I. Jordan (29:12.880)

Just like anyone who ran a big auction or ran a big matching service in economics will

Michael I. Jordan (29:17.240)

think these things through.

Lex Fridman (29:18.880)

And so that maybe doesn't get at all the huge issues that can arise when you start to create

Michael I. Jordan (29:22.520)

markets, but it starts to, at least for me, solidify my thoughts and allow me to move

Lex Fridman (29:26.760)

forward in my own thinking.

Michael I. Jordan (29:28.080)

Yeah.

Lex Fridman (29:29.080)

So I talked to the head of research at Spotify actually, and I think their longterm goal,

Michael I. Jordan (29:32.840)

they've said, is to have at least one million creators make a comfortable living putting

Lex Fridman (29:39.920)

on Spotify.

Lex Fridman (29:41.120)

So I think you articulate a really nice vision of the world and the digital and the cyberspace

Lex Fridman (29:52.160)

of markets.

Lex Fridman (29:53.920)

What do you think companies like Spotify or YouTube or Netflix can do to create such markets?

Lex Fridman (30:04.100)

Is it an AI problem?

Lex Fridman (30:05.400)

Is it an interface problem for interface design?

Lex Fridman (30:08.600)

Is it some other kind of, is it an economics problem?

Lex Fridman (30:13.440)

Who should they hire to solve these problems?

Lex Fridman (30:15.600)

Well, part of it's not just top down.

Lex Fridman (30:17.480)

So the Silicon Valley has this attitude that they know how to do it.

Michael I. Jordan (30:20.000)

They will create the system just like Google did with the search box that will be so good

Michael I. Jordan (30:23.380)

that they'll just, everyone will adopt that.

Lex Fridman (30:27.000)

It's everything you said, but really I think missing that kind of culture.

Lex Fridman (30:31.500)

So it's literally that 16 year old who's able to create the songs.

Lex Fridman (30:34.800)

You don't create that as a Silicon Valley entity.

Michael I. Jordan (30:37.000)

You don't hire them per se.

Lex Fridman (30:39.340)

You have to create an ecosystem in which they are wanted and that they belong.

Lex Fridman (30:44.320)

And so you have to have some cultural credibility to do things like this.

Michael I. Jordan (30:47.680)

Netflix, to their credit, wanted some of that credibility and they created shows, content.

Michael I. Jordan (30:53.060)

They call it content.

Lex Fridman (30:54.060)

It's such a terrible word, but it's culture.

Lex Fridman (30:56.880)

And so with movies, you can kind of go give a large sum of money to somebody graduating

Lex Fridman (31:01.160)

from the USC film school.

Michael I. Jordan (31:03.440)

It's a whole thing of its own, but it's kind of like rich white people's thing to do.

Lex Fridman (31:07.880)

And American culture has not been so much about rich white people.

Michael I. Jordan (31:11.760)

It's been about all the immigrants, all the Africans who came and brought that culture

Lex Fridman (31:16.580)

and those rhythms to this world and created this whole new thing.

Michael I. Jordan (31:23.040)

American culture.

Lex Fridman (31:24.040)

And so companies can't artificially create that.

Michael I. Jordan (31:26.800)

They can't just say, hey, we're here.

Lex Fridman (31:28.440)

We're going to buy it up.

Michael I. Jordan (31:29.440)

You've got a partner.

Lex Fridman (31:31.440)

And so anyway, not to denigrate, these companies are all trying and they should, and I'm sure

Michael I. Jordan (31:37.520)

they're asking these questions and some of them are even making an effort.

Lex Fridman (31:40.160)

But it is partly a respect the culture as a technology person.

Michael I. Jordan (31:44.400)

You've got to blend your technology with cultural meaning.

Lex Fridman (31:49.880)

How much of a role do you think the algorithm, so machine learning has in connecting the

Lex Fridman (31:54.400)

consumer to the creator, sort of the recommender system aspect of this?

Lex Fridman (31:59.600)

Yeah.

Michael I. Jordan (32:00.600)

It's a great question.

Lex Fridman (32:01.600)

I think pretty high.

Michael I. Jordan (32:02.600)

There's no magic in the algorithms, but a good recommender system is way better than

Lex Fridman (32:07.320)

a bad recommender system.

Lex Fridman (32:09.160)

And recommender systems is a billion dollar industry back even 10, 20 years ago.

Lex Fridman (32:15.180)

And it continues to be extremely important going forward.

Michael I. Jordan (32:17.540)

What's your favorite recommender system, just so we can put something, well, just historically

Michael I. Jordan (32:20.680)

I was one of the, when I first went to Amazon, I first didn't like Amazon because they put

Michael I. Jordan (32:24.800)

the book people out of business or the library, the local booksellers went out of business.

Michael I. Jordan (32:30.400)

I've come to accept that there probably are more books being sold now and poor people

Michael I. Jordan (32:34.620)

reading them than ever before.

Lex Fridman (32:36.920)

And then local book stores are coming back.

Lex Fridman (32:39.440)

So that's how economics sometimes work.

Lex Fridman (32:41.540)

You go up and you go down.

Lex Fridman (32:44.280)

But anyway, when I finally started going there and I bought a few books, I was really pleased

Michael I. Jordan (32:48.760)

to see another few books being recommended to me that I never would have thought of.

Lex Fridman (32:52.400)

And I bought a bunch of them.

Lex Fridman (32:53.400)

So they obviously had a good business model.

Lex Fridman (32:55.320)

But I learned things and I still to this day kind of browse using that service.

Lex Fridman (33:00.980)

And I think lots of people get a lot, that is a good aspect of a recommendation system.

Michael I. Jordan (33:05.760)

I'm learning from my peers in an indirect way.

Lex Fridman (33:10.480)

And their algorithms are not meant to have them impose what we learn.

Michael I. Jordan (33:13.880)

It really is trying to find out what's in the data.

Michael I. Jordan (33:16.680)

It doesn't work so well for other kinds of entities, but that's just the complexity of

Michael I. Jordan (33:19.680)

human life.

Michael I. Jordan (33:20.680)

Like shirts, I'm not going to get recommendations on shirts, but that's interesting.

Michael I. Jordan (33:26.440)

If you try to recommend restaurants, it's hard.

Lex Fridman (33:32.160)

It's hard to do it at scale.

Lex Fridman (33:35.400)

But a blend of recommendation systems with other economic ideas, matchings and so on

Lex Fridman (33:42.080)

is really, really still very open research wise.

Lex Fridman (33:45.240)

And there's new companies that are going to emerge that do that well.

Lex Fridman (33:48.680)

What do you think is going to the messy, difficult land of say politics and things like that,

Lex Fridman (33:54.480)

that YouTube and Twitter have to deal with in terms of recommendation systems?

Lex Fridman (33:58.480)

Being able to suggest, I think Facebook just launched Facebook news.

Lex Fridman (34:03.120)

So recommend the kind of news that are most likely for you to be interesting.

Lex Fridman (34:08.920)

Do you think this is AI solvable, again, whatever term we want to use, do you think it's a solvable

Lex Fridman (34:14.520)

problem for machines or is it a deeply human problem that's unsolvable?

Lex Fridman (34:18.760)

So I don't even think about it at that level.

Michael I. Jordan (34:20.240)

I think that what's broken with some of these companies, it's all monetization by advertising.

Michael I. Jordan (34:25.400)

They're not, at least Facebook, I want to critique them, but they didn't really try

Lex Fridman (34:29.200)

to connect a producer and a consumer in an economic way, right?

Lex Fridman (34:32.680)

No one wants to pay for anything.

Lex Fridman (34:34.700)

And so they all, you know, starting with Google and Facebook, they went back to the playbook

Lex Fridman (34:38.420)

of, you know, the television companies back in the day.

Michael I. Jordan (34:41.440)

No one wanted to pay for this signal.

Lex Fridman (34:43.200)

They will pay for the TV box, but not for the signal, at least back in the day.

Lex Fridman (34:47.200)

And so advertising kind of filled that gap and advertising was new and interesting and

Lex Fridman (34:50.400)

it somehow didn't take over our lives quite, right?

Michael I. Jordan (34:54.400)

Fast forward, Google provides a service that people don't want to pay for.

Lex Fridman (34:59.880)

And so somewhat surprisingly in the nineties, they made, they ended up making huge amounts

Lex Fridman (35:04.120)

so they cornered the advertising market.

Lex Fridman (35:05.600)

It didn't seem like that was going to happen, at least to me.

Michael I. Jordan (35:08.400)

These little things on the right hand side of the screen just did not seem all that economically

Lex Fridman (35:11.720)

interesting, but that companies had maybe no other choice.

Michael I. Jordan (35:14.360)

The TV market was going away and billboards and so on.

Lex Fridman (35:17.800)

So they've, they got it.

Lex Fridman (35:19.860)

And I think that sadly that Google just has, it was doing so well with that at making such

Lex Fridman (35:24.880)

money.

Michael I. Jordan (35:25.880)

They didn't think much more about how, wait a minute, is there a producer consumer relationship

Lex Fridman (35:28.700)

to be set up here?

Michael I. Jordan (35:29.700)

Not just between us and the advertisers market to be created.

Lex Fridman (35:32.840)

Is there an actual market between the producer consumer?

Michael I. Jordan (35:35.160)

They're the producers, the person who created that video clip, the person that made that

Michael I. Jordan (35:38.240)

website, the person who could make more such things, the person who could adjust it as

Michael I. Jordan (35:42.000)

a function of demand, the person on the other side who's asking for different kinds of things,

Lex Fridman (35:46.800)

you know?

Lex Fridman (35:47.800)

So you see glimmers of that now there's influencers and there's kind of a little glimmering of

Lex Fridman (35:51.320)

a market, but it should have been done 20 years ago.

Michael I. Jordan (35:53.480)

It should have been thought about.

Lex Fridman (35:54.480)

It should have been created in parallel with the advertising ecosystem.

Lex Fridman (35:58.400)

And then Facebook inherited that.

Lex Fridman (35:59.860)

And I think they also didn't think very much about that.

Lex Fridman (36:03.160)

So fast forward and now they are making huge amounts of money off of advertising.

Lex Fridman (36:07.960)

And the news thing and all these clicks is just feeding the advertising.

Michael I. Jordan (36:11.560)

It's all connected up to the advertiser.

Lex Fridman (36:13.640)

So you want more people to click on certain things because that money flows to you, Facebook.

Michael I. Jordan (36:18.580)

You're very much incentivized to do that.

Lex Fridman (36:20.000)

And when you start to find it's breaking, people are telling you, well, we're getting

Michael I. Jordan (36:23.480)

into some troubles.

Lex Fridman (36:24.480)

You try to adjust it with your smart AI algorithms, right?

Lex Fridman (36:27.580)

And figure out what are bad clicks.

Lex Fridman (36:28.960)

So maybe it shouldn't be click through rate, it should be something else.

Michael I. Jordan (36:31.040)

I find that pretty much hopeless.

Lex Fridman (36:34.080)

It does get into all the complexity of human life and you can try to fix it.

Michael I. Jordan (36:37.400)

You should, but you could also fix the whole business model.

Lex Fridman (36:40.840)

And the business model is that really, what are, are there some human producers and consumers

Lex Fridman (36:44.600)

out there?

Lex Fridman (36:45.600)

Is there some economic value to be liberated by connecting them directly?

Lex Fridman (36:48.760)

Is it such that it's so valuable that people will be able to pay for it?

Lex Fridman (36:52.640)

All right.

Lex Fridman (36:53.640)

And micro payments, like small payments.

Lex Fridman (36:54.640)

Micro, but even have to be micro.

Lex Fridman (36:56.620)

So I like the example, suppose I'm going, next week I'm going to India.

Lex Fridman (37:00.120)

Never been to India before.

Lex Fridman (37:01.120)

Right?

Lex Fridman (37:02.120)

I have a couple of days in Mumbai, I have no idea what to do there.

Lex Fridman (37:06.560)

Right?

Lex Fridman (37:07.560)

And I could go on the web right now and search.

Michael I. Jordan (37:08.880)

It's going to be kind of hopeless.

Lex Fridman (37:10.080)

I'm not going to find, you know, I have lots of advertisers in my face.

Lex Fridman (37:14.080)

Right?

Lex Fridman (37:15.080)

What I really want to do is broadcast to the world that I am going to Mumbai and have someone

Michael I. Jordan (37:19.320)

on the other side of a market look at me and, and there's a recommendation system there.

Lex Fridman (37:24.000)

So I'm not looking at all possible people coming to Mumbai.

Michael I. Jordan (37:26.040)

They're looking at the people who are relevant to them.

Lex Fridman (37:27.680)

So someone in my age group, someone who kind of knows me in some level, I give up a little

Michael I. Jordan (37:32.480)

privacy by that, but I'm happy because what I'm going to get back is this person can make

Michael I. Jordan (37:35.720)

a little video for me, or they're going to write a little two page paper on here's the

Lex Fridman (37:39.320)

cool things that you want to do and move by this week, especially, right?

Lex Fridman (37:43.160)

I'm going to look at that.

Michael I. Jordan (37:44.160)

I'm not going to pay a micro payment.

Lex Fridman (37:45.160)

I'm going to pay, you know, a hundred dollars or whatever for that.

Michael I. Jordan (37:48.000)

It's real value.

Lex Fridman (37:49.000)

It's like journalism.

Michael I. Jordan (37:50.000)

Um, and as an honest subscription, it's that I'm going to pay that person in that moment.

Lex Fridman (37:54.920)

Company's going to take 5% of that.

Lex Fridman (37:56.680)

And that person has now got it.

Michael I. Jordan (37:57.760)

It's a gig economy, if you will, but you know, done for it, you know, thinking about a little

Michael I. Jordan (38:01.240)

bit behind YouTube, there was actually people who could make more of those things.

Michael I. Jordan (38:05.000)

If they were connected to a market, they would make more of those things independently.

Michael I. Jordan (38:07.960)

You don't have to tell them what to do.

Lex Fridman (38:08.960)

You don't have to incentivize them any other way.

Michael I. Jordan (38:11.680)

Um, and so, yeah, these companies, I don't think have thought long and hard about that.

Lex Fridman (38:15.760)

So I do distinguish on Facebook on the one side, who just not thought about these things

Michael I. Jordan (38:20.160)

at all.

Michael I. Jordan (38:21.160)

I think, uh, thinking that AI will fix everything, uh, and Amazon thinks about them all the time

Michael I. Jordan (38:25.200)

because they were already out in the real world.

Lex Fridman (38:26.520)

They were delivering packages, people's doors.

Michael I. Jordan (38:28.080)

They were, they were worried about a market.

Michael I. Jordan (38:29.400)

They were worried about sellers and, you know, they worry and some things they do are great.

Michael I. Jordan (38:32.600)

Some things maybe not so great, but you know, they're in that business model.

Lex Fridman (38:36.440)

And then I'd say Google sort of hovers somewhere in between.

Michael I. Jordan (38:38.360)

I don't, I don't think for a long, long time they got it.

Michael I. Jordan (38:41.440)

I think they probably see that YouTube is more pregnant with possibility than, than,

Michael I. Jordan (38:45.720)

than they might've thought and that they're probably heading that direction.

Michael I. Jordan (38:49.120)

Um, but uh, you know, Silicon Valley has been dominated by the Google Facebook kind of mentality

Lex Fridman (38:54.000)

and the subscription and advertising and that is, that's the core problem, right?

Michael I. Jordan (38:58.800)

The fake news actually rides on top of that because it means that you're monetizing with

Michael I. Jordan (39:03.640)

clip through rate and that is the core problem.

Lex Fridman (39:05.600)

You got to remove that.

Lex Fridman (39:06.880)

So advertisement, if we're going to linger on that, I mean, that's an interesting thesis.

Lex Fridman (39:11.200)

I don't know if everyone really deeply thinks about that.

Lex Fridman (39:15.060)

So you're right.

Michael I. Jordan (39:16.720)

The thought is the advertising model is the only thing we have, the only thing we'll ever

Michael I. Jordan (39:20.960)

have.

Michael I. Jordan (39:21.960)

We have to fix, we have to build algorithms that despite that business model, you know,

Michael I. Jordan (39:30.240)

find the better angels of our nature and do good by society and by the individual.

Lex Fridman (39:34.680)

But you think we can slowly, you think, first of all, there's a difference between should

Lex Fridman (39:40.000)

and could.

Lex Fridman (39:42.040)

So you're saying we should slowly move away from the advertising model and have a direct

Michael I. Jordan (39:46.600)

connection between the consumer and the creator.

Michael I. Jordan (39:49.920)

The question I also have is, can we, because the advertising model is so successful now

Michael I. Jordan (39:55.240)

in terms of just making a huge amount of money and therefore being able to build a big company

Lex Fridman (40:00.400)

that provides, has really smart people working that create a good service.

Lex Fridman (40:03.920)

Do you think it's possible?

Lex Fridman (40:05.680)

And just to clarify, you think we should move away?

Michael I. Jordan (40:07.880)

Well, I think we should.

Lex Fridman (40:08.880)

Yeah.

Lex Fridman (40:09.880)

But we is the, you know, me.

Lex Fridman (40:10.880)

So society.

Michael I. Jordan (40:11.880)

Yeah.

Michael I. Jordan (40:12.880)

Well, the companies, I mean, so first of all, full disclosure, I'm doing a day a week at

Michael I. Jordan (40:16.360)

Amazon because I kind of want to learn more about how they do things.

Michael I. Jordan (40:18.840)

So, you know, I'm not speaking for Amazon in any way, but, you know, I did go there

Michael I. Jordan (40:22.760)

because I actually believe they get a little bit of this or trying to create these markets.

Lex Fridman (40:26.240)

And they don't really use, advertising is not a crucial part of it.

Michael I. Jordan (40:29.520)

Well, that's a good question.

Lex Fridman (40:30.520)

So it has become not crucial, but it's become more and more present if you go to Amazon

Michael I. Jordan (40:34.840)

website.

Michael I. Jordan (40:35.840)

And, you know, without revealing too many deep secrets about Amazon, I can tell you

Michael I. Jordan (40:38.840)

that, you know, a lot of people in the company question this and there's a huge questioning

Lex Fridman (40:42.480)

going on.

Michael I. Jordan (40:43.620)

You do not want a world where there's zero advertising.

Lex Fridman (40:45.660)

That actually is a bad world.

Michael I. Jordan (40:47.160)

Okay.

Lex Fridman (40:48.160)

So here's a way to think about it.

Lex Fridman (40:49.280)

You're a company that like Amazon is trying to bring products to customers, right?

Lex Fridman (40:55.000)

And the customer, at any given moment, you want to buy a vacuum cleaner, say, you want

Michael I. Jordan (40:58.360)

to know what's available for me.

Lex Fridman (40:59.360)

And, you know, it's not going to be that obvious.

Michael I. Jordan (41:00.840)

You have to do a little bit of work at it.

Lex Fridman (41:02.160)

The recommendation system will sort of help, right?

Lex Fridman (41:04.600)

But now suppose this other person over here has just made the world, you know, they spent

Lex Fridman (41:08.080)

a huge amount of energy.

Michael I. Jordan (41:09.080)

They had a great idea.

Lex Fridman (41:10.080)

They made a great vacuum cleaner.

Michael I. Jordan (41:11.080)

They know they really did it.

Lex Fridman (41:12.400)

They nailed it.

Lex Fridman (41:13.400)

It's an MIT, you know, whiz kid that made a great new vacuum cleaner, right?

Lex Fridman (41:16.680)

It's not going to be in the recommendation system.

Michael I. Jordan (41:18.240)

No one will know about it.

Lex Fridman (41:19.280)

The algorithms will not find it and AI will not fix that.

Michael I. Jordan (41:22.440)

Okay.

Lex Fridman (41:23.440)

At all.

Michael I. Jordan (41:24.440)

Right.

Lex Fridman (41:25.440)

How do you allow that vacuum cleaner to start to get in front of people, be sold well advertising.

Lex Fridman (41:30.660)

And here, what advertising is, it's a signal that you're, you believe in your product enough

Lex Fridman (41:35.360)

that you're willing to pay some real money for it.

Lex Fridman (41:37.480)

And to me as a consumer, I look at that signal.

Michael I. Jordan (41:39.480)

I say, well, first of all, I know these are not just cheap little ads cause we have now

Michael I. Jordan (41:43.240)

right now there.

Lex Fridman (41:44.240)

I know that, you know, these are super cheap, you know, pennies.

Michael I. Jordan (41:47.740)

If I see an ad where it's actually, I know the company is only doing a few of these and

Michael I. Jordan (41:51.120)

they're making, you know, real money is kind of flowing and I see an ad, I may pay more

Michael I. Jordan (41:54.520)

attention to it.

Lex Fridman (41:55.520)

And I actually might want that because I see, Hey, that guy spent money on his vacuum cleaner.

Michael I. Jordan (42:01.600)

Maybe there's something good there.

Lex Fridman (42:02.600)

So I will look at it.

Lex Fridman (42:03.600)

And so that's part of the overall information flow in a good market.

Lex Fridman (42:06.620)

So advertising has a role, but the problem is of course that that signal is now completely

Michael I. Jordan (42:11.720)

gone because it just, you know, dominant by these tiny little things that add up to big

Lex Fridman (42:15.800)

money for the company, you know?

Lex Fridman (42:17.740)

So I think it will just, I think it will change because the societies just don't, you know,

Michael I. Jordan (42:22.600)

stick with things that annoy a lot of people and advertising currently annoys people more

Michael I. Jordan (42:26.480)

than it provides information.

Lex Fridman (42:28.480)

And I think that a Google probably is smart enough to figure out that this is a dead,

Michael I. Jordan (42:32.200)

this is a bad model, even though it's a hard, huge amount of money and they'll have to figure

Lex Fridman (42:35.760)

out how to pull it away from it slowly.

Lex Fridman (42:38.080)

And I'm sure the CEO there will figure it out, but they need to do it.

Lex Fridman (42:42.280)

And they needed it to, so if you reduce advertising, not to zero, but you reduce it at the same

Michael I. Jordan (42:47.120)

time you bring up producer, consumer, actual real value being delivered.

Lex Fridman (42:51.640)

So real money is being paid and they take a 5% cut that 5% could start to get big enough

Michael I. Jordan (42:56.260)

to cancel out the lost revenue from the kind of the poor kind of advertising.

Lex Fridman (43:00.080)

And I think that a good company will do that, will realize that.

Lex Fridman (43:04.740)

And Facebook, you know, again, God bless them.

Michael I. Jordan (43:08.440)

They bring, you know, grandmothers, they bring children's pictures into grandmothers lives.

Michael I. Jordan (43:14.680)

It's fantastic.

Lex Fridman (43:17.340)

But they need to think of a new business model and that's the core problem there.

Michael I. Jordan (43:22.440)

Until they start to connect producer consumer, I think they will just continue to make money

Lex Fridman (43:26.440)

and then buy the next social network company and then buy the next one and the innovation

Michael I. Jordan (43:30.560)

level will not be high and the health issues will not go away.

Lex Fridman (43:34.880)

So I apologize that we kind of returned to words, I don't think the exact terms matter,

Lex Fridman (43:41.120)

but in sort of defense of advertisement, don't you think the kind of direct connection between

Lex Fridman (43:49.440)

consumer and creator producer is what advertisement strives to do, right?

Lex Fridman (44:00.960)

So that is best advertisement is literally now Facebook is listening to our conversation

Lex Fridman (44:06.680)

and heard that you're going to India and will be able to actually start automatically for

Michael I. Jordan (44:11.400)

you making these connections and start giving this offer.

Lex Fridman (44:14.500)

So like, I apologize if it's just a matter of terms, but just to draw a distinction,

Michael I. Jordan (44:19.800)

is it possible to make advertisements just better and better and better algorithmically

Lex Fridman (44:23.000)

to where it actually becomes a connection, almost a direct connection?

Michael I. Jordan (44:26.040)

That's a good question.

Lex Fridman (44:27.040)

So let's component on that.

Michael I. Jordan (44:28.040)

First of all, what we just talked about, I was defending advertising.

Lex Fridman (44:32.000)

Okay.

Lex Fridman (44:33.000)

So I was defending it as a way to get signals into a market that don't come any other way,

Lex Fridman (44:36.400)

especially algorithmically.

Michael I. Jordan (44:37.720)

It's a sign that someone spent money on it, it's a sign they think it's valuable.

Lex Fridman (44:41.640)

And if I think that if other things, someone else thinks it's valuable, and if I trust

Michael I. Jordan (44:45.020)

other people, I might be willing to listen.

Lex Fridman (44:47.360)

I don't trust that Facebook though, who's an intermediary between this.

Michael I. Jordan (44:51.840)

I don't think they care about me.

Lex Fridman (44:54.600)

Okay.

Michael I. Jordan (44:55.600)

I don't think they do.

Lex Fridman (44:56.720)

And I find it creepy that they know I'm going to India next week because of our conversation.

Lex Fridman (45:00.880)

Why do you think that is?

Lex Fridman (45:02.360)

So what, could you just put your PR hat on?

Lex Fridman (45:07.120)

Why do you think you find Facebook creepy and not trust them as do majority of the population?

Lex Fridman (45:14.180)

So they're out of the Silicon Valley companies, I saw like not approval rate, but there's

Michael I. Jordan (45:19.360)

ranking of how much people trust companies and Facebook is in the gutter.

Lex Fridman (45:23.080)

In the gutter, including people inside of Facebook.

Lex Fridman (45:25.600)

So what do you attribute that to?

Lex Fridman (45:28.000)

Because when I...

Michael I. Jordan (45:29.000)

Come on, you don't find it creepy that right now we're talking that I might walk out on

Michael I. Jordan (45:31.840)

the street right now that some unknown person who I don't know kind of comes up to me and

Michael I. Jordan (45:35.880)

says, I hear you're going to India.

Lex Fridman (45:37.500)

I mean, that's not even Facebook.

Michael I. Jordan (45:38.900)

That's just, I want transparency in human society.

Michael I. Jordan (45:42.560)

I want to have, if you know something about me, there's actually some reason you know

Michael I. Jordan (45:45.680)

something about me.

Lex Fridman (45:47.080)

That's something that if I look at it later and audit it kind of, I approve.

Michael I. Jordan (45:51.560)

You know something about me because you care in some way.

Lex Fridman (45:54.580)

There's a caring relationship even, or an economic one or something.

Michael I. Jordan (45:58.240)

Not just that you're someone who could exploit it in ways I don't know about or care about

Lex Fridman (46:02.000)

or I'm troubled by or whatever.

Michael I. Jordan (46:05.240)

We're in a world right now where that happens way too much and that Facebook knows things

Lex Fridman (46:09.880)

about a lot of people and could exploit it and does exploit it at times.

Michael I. Jordan (46:14.720)

I think most people do find that creepy.

Lex Fridman (46:16.880)

It's not for them.

Michael I. Jordan (46:17.880)

It's not that Facebook is not doing it because they care about them in a real sense.

Lex Fridman (46:23.440)

And they shouldn't.

Michael I. Jordan (46:24.440)

They should not be a big brother caring about us.

Lex Fridman (46:26.740)

That is not the role of a company like that.

Lex Fridman (46:28.560)

Why not?

Lex Fridman (46:29.560)

Wait, not the big brother part, but the caring, the trusting.

Michael I. Jordan (46:32.160)

I mean, don't those companies, just to link on it because a lot of companies have a lot

Lex Fridman (46:37.120)

of information about us.

Michael I. Jordan (46:38.320)

I would argue that there's companies like Microsoft that has more information about

Lex Fridman (46:42.560)

us than Facebook does and yet we trust Microsoft more.

Michael I. Jordan (46:46.000)

Well, Microsoft is pivoting.

Lex Fridman (46:47.480)

Microsoft, you know, under Satya Nadella has decided this is really important.

Michael I. Jordan (46:51.360)

We don't want to do creepy things.

Michael I. Jordan (46:53.320)

Really want people to trust us to actually only use information in ways that they really

Lex Fridman (46:56.720)

would approve of, that we don't decide, right?

Lex Fridman (47:00.360)

And I'm just kind of adding that the health of a market is that when I connect to someone

Michael I. Jordan (47:06.640)

who produces a consumer, it's not just a random producer or consumer, it's people who see

Lex Fridman (47:10.160)

each other.

Michael I. Jordan (47:11.160)

They don't like each other, but they sense that if they transact, some happiness will

Lex Fridman (47:14.360)

go up on both sides.

Michael I. Jordan (47:15.940)

If a company helps me to do that in moments that I choose of my choosing, then fine.

Lex Fridman (47:22.800)

So, and also think about the difference between, you know, browsing versus buying, right?

Michael I. Jordan (47:28.560)

There are moments in my life I just want to buy, you know, a gadget or something.

Lex Fridman (47:31.760)

I need something for that moment.

Michael I. Jordan (47:33.080)

I need some ammonia for my house or something because I got a problem with a spill.

Lex Fridman (47:37.400)

I want to just go in.

Michael I. Jordan (47:38.400)

I don't want to be advertised at that moment.

Lex Fridman (47:40.080)

I don't want to be led down various, you know, that's annoying.

Michael I. Jordan (47:43.040)

I want to just go and have it be extremely easy to do what I want.

Lex Fridman (47:49.020)

Other moments I might say, no, it's like today I'm going to the shopping mall.

Michael I. Jordan (47:52.440)

I want to walk around and see things and see people and be exposed to stuff.

Lex Fridman (47:55.560)

So I want control over that though.

Lex Fridman (47:56.800)

I don't want the company's algorithms to decide for me, right?

Lex Fridman (48:00.200)

I think that's the thing.

Michael I. Jordan (48:01.200)

There's a total loss of control if Facebook thinks they should take the control from us

Michael I. Jordan (48:04.880)

of deciding when we want to have certain kinds of information, when we don't, what information

Michael I. Jordan (48:08.200)

that is, how much it relates to what they know about us that we didn't really want them

Lex Fridman (48:11.880)

to know about us.

Michael I. Jordan (48:13.680)

I don't want them to be helping me in that way.

Michael I. Jordan (48:15.840)

I don't want them to be helping them by they decide they have control over what I want

Lex Fridman (48:21.640)

and when.

Lex Fridman (48:22.640)

I totally agree.

Michael I. Jordan (48:23.640)

Facebook, by the way, I have this optimistic thing where I think Facebook has the kind

Lex Fridman (48:28.560)

of personal information about us that could create a beautiful thing.

Lex Fridman (48:32.480)

So I'm really optimistic of what Facebook could do.

Lex Fridman (48:36.200)

It's not what it's doing, but what it could do.

Lex Fridman (48:38.680)

So I don't see that.

Michael I. Jordan (48:39.840)

I think that optimism is misplaced because there's not a bit, you have to have a business

Michael I. Jordan (48:43.400)

model behind these things.

Lex Fridman (48:44.400)

Create a beautiful thing is really, let's be, let's be clear.

Michael I. Jordan (48:48.480)

It's about something that people would value.

Lex Fridman (48:51.400)

And I don't think they have that business model and I don't think they will suddenly

Michael I. Jordan (48:55.080)

discover it by what, you know, a long hot shower.

Lex Fridman (48:58.920)

I disagree.

Michael I. Jordan (48:59.920)

I disagree in terms of, you can discover a lot of amazing things in a shower.

Lex Fridman (49:04.840)

So I didn't say that.

Michael I. Jordan (49:05.840)

I said, they won't come, they won't do it, but in the shower, I think a lot of other

Lex Fridman (49:10.240)

people will discover it.

Michael I. Jordan (49:11.300)

I think that this guy, so I should also, full disclosure, there's a company called United

Michael I. Jordan (49:15.240)

Masters, which I'm on their board and they've created this music market and I have a hundred

Michael I. Jordan (49:18.760)

thousand artists now signed on and they've done things like gone to the NBA and the NBA,

Lex Fridman (49:23.220)

the music you find behind NBA clips right now is their music, right?

Lex Fridman (49:26.960)

That's a company that had the right business model in mind from the get go, right?

Lex Fridman (49:31.920)

Executed on that.

Lex Fridman (49:32.920)

And from day one, there was value brought to, so here you have a kid who made some songs

Lex Fridman (49:37.220)

who suddenly their songs are on the NBA website, right?

Michael I. Jordan (49:41.260)

That's real economic value to people.

Lex Fridman (49:43.440)

And so, you know, so you and I differ on the optimism of being able to sort of change the

Lex Fridman (49:51.800)

direction of the Titanic, right?

Lex Fridman (49:54.440)

So I, yeah, I'm older than you, so I've seen some Titanic's crash, got it.

Lex Fridman (50:01.120)

But and just to elaborate, cause I totally agree with you and I just want to know how

Michael I. Jordan (50:05.560)

difficult you think this problem is of, so for example, I want to read some news and

Michael I. Jordan (50:11.880)

I would, there's a lot of times in the day where something makes me either smile or think

Lex Fridman (50:16.940)

in a way where I like consciously think this really gave me value.

Michael I. Jordan (50:20.800)

Like I sometimes listen to the daily podcasts in the New York times, way better than the

Lex Fridman (50:26.480)

New York times themselves, by the way, for people listening.

Michael I. Jordan (50:29.320)

That's like real journalism is happening for some reason in the podcast space.

Michael I. Jordan (50:32.560)

It doesn't make sense to me, but often I listen to it 20 minutes and I would be willing to

Michael I. Jordan (50:37.600)

pay for that, like $5, $10 for that experience.

Lex Fridman (50:41.860)

And how difficult, that's kind of what you're getting at is that little transaction.

Lex Fridman (50:48.200)

How difficult is it to create a frictionless system like Uber has, for example, for other

Lex Fridman (50:52.640)

things?

Lex Fridman (50:53.640)

What's your intuition there?

Lex Fridman (50:55.280)

So I, first of all, I pay little bits of money to, you know, to send, there's something

Michael I. Jordan (50:58.500)

called courts that does financial things.

Lex Fridman (51:00.300)

I like medium as a site, I don't pay there, but I would.

Michael I. Jordan (51:04.480)

You had a great post on medium.

Lex Fridman (51:06.280)

I would have loved to pay you a dollar and not others.

Michael I. Jordan (51:10.280)

I wouldn't have wanted it per se because there should be also sites where that's not actually

Lex Fridman (51:15.560)

the goal.

Michael I. Jordan (51:16.560)

The goal is to actually have a broadcast channel that I monetize in some other way if I chose

Lex Fridman (51:20.240)

to.

Michael I. Jordan (51:21.240)

I mean, I could now people know about it.

Lex Fridman (51:23.080)

I could, I'm not doing it, but that's fine with me.

Michael I. Jordan (51:26.360)

Also the musicians who are making all this music, I don't think the right model is that

Lex Fridman (51:29.840)

you pay a little subscription fee to them, right?

Michael I. Jordan (51:32.880)

Because people can copy the bits too easily and it's just not that somewhere the value

Lex Fridman (51:35.860)

is.

Michael I. Jordan (51:36.860)

The value is that a connection was made between real human beings, then you can follow up

Lex Fridman (51:39.800)

on that.

Michael I. Jordan (51:40.800)

All right.

Lex Fridman (51:41.800)

And create yet more value.

Lex Fridman (51:42.960)

So no, I think there's a lot of open questions here, hot open questions, but also, yeah,

Lex Fridman (51:47.920)

I do want good recommendation systems that recommend cool stuff to me.

Lex Fridman (51:51.360)

But it's pretty hard, right?

Lex Fridman (51:52.360)

I don't like them to recommend stuff just based on my browsing history.

Michael I. Jordan (51:55.880)

I don't like the based on stuff they know about me, quote unquote.

Lex Fridman (51:59.000)

What's unknown about me is the most interesting.

Lex Fridman (52:00.860)

So this is the, this is the really interesting question.

Lex Fridman (52:03.640)

We may disagree, maybe not.

Michael I. Jordan (52:05.860)

I think that I love recommender systems and I want to give them everything about me in

Lex Fridman (52:12.160)

a way that I trust.

Michael I. Jordan (52:13.160)

Yeah.

Lex Fridman (52:14.160)

But you, but you don't, because, so for example, this morning I clicked on a, you know, I was

Michael I. Jordan (52:17.880)

pretty sleepy this morning.

Lex Fridman (52:19.960)

I clicked on a story about the queen of England.

Michael I. Jordan (52:23.280)

Yes.

Lex Fridman (52:24.280)

Right.

Michael I. Jordan (52:25.280)

I do not give a damn about the queen of England.

Lex Fridman (52:26.440)

I really do not.

Lex Fridman (52:27.560)

But it was clickbait.

Lex Fridman (52:28.560)

It kind of looked funny and I had to say, what the heck are they talking about?

Michael I. Jordan (52:31.520)

I don't want to have my life, you know, heading that direction.

Lex Fridman (52:34.040)

Now that's in my browsing history.

Michael I. Jordan (52:36.180)

The system in any reasonable system will think that I care about the queen of England.

Lex Fridman (52:39.880)

That's browsing history.

Michael I. Jordan (52:40.880)

Right.

Michael I. Jordan (52:41.880)

But, but you're saying all the trace, all the digital exhaust or whatever, that's been

Michael I. Jordan (52:44.640)

kind of the models.

Lex Fridman (52:45.640)

If you collect all this stuff, you're going to figure all of us out.

Michael I. Jordan (52:48.560)

Well, if you're trying to figure out like kind of one person like Trump or something,

Lex Fridman (52:51.280)

maybe you could figure him out.

Lex Fridman (52:52.280)

But if you're trying to figure out, you know, 500 million people, you know, no way, no way.

Lex Fridman (52:58.040)

You think so?

Michael I. Jordan (52:59.040)

No, I do.

Lex Fridman (53:00.040)

I think so.

Michael I. Jordan (53:01.040)

I think we are, humans are just amazingly rich and complicated.

Michael I. Jordan (53:02.560)

Every one of us has our little quirks, every one of us has our little things that could

Michael I. Jordan (53:05.220)

intrigue us that we don't even know it will intrigue us.

Lex Fridman (53:08.020)

And there's no sign of it in our past, but by God, there it comes and you know, you fall

Michael I. Jordan (53:12.240)

in love with it.

Lex Fridman (53:13.240)

And I don't want a company trying to figure that out for me and anticipate that I want

Michael I. Jordan (53:16.520)

them to provide a forum, a market, a place that I kind of go and by hook or by crook,

Michael I. Jordan (53:22.160)

this happens, you know, I I'm walking down the street and I hear some Chilean music being

Michael I. Jordan (53:26.120)

played and I never knew I liked Chilean music, but wow.

Lex Fridman (53:28.580)

So there is that side and I want them to provide a limited, but you know, interesting place

Michael I. Jordan (53:33.680)

to go.

Lex Fridman (53:34.680)

Right.

Lex Fridman (53:35.680)

And so don't try to use your AI to kind of, you know, figure me out and then put me in

Michael I. Jordan (53:39.740)

a world where you figured me out, you know, no, create huge spaces for human beings where

Michael I. Jordan (53:45.140)

our creativity and our style will be enriched and come forward and it'll be a lot of more

Lex Fridman (53:50.360)

transparency.

Michael I. Jordan (53:51.360)

I won't have people randomly, anonymously putting comments up and I'll special based

Lex Fridman (53:55.400)

on stuff they know about me, facts that, you know, we are so broken right now.

Michael I. Jordan (54:00.080)

If you're, you know, especially if you're a celebrity, but you know, it's about anybody

Lex Fridman (54:02.920)

that anonymous people are hurting lots and lots of people right now.

Michael I. Jordan (54:06.720)

That's part of this thing that Silicon Valley is thinking that, you know, just collect all

Lex Fridman (54:10.200)

this information and use it in a great way.

Lex Fridman (54:12.480)

So no, I'm not, I'm not a pessimist, I'm very much an optimist by nature, but I think that's

Lex Fridman (54:16.420)

just been the wrong path for the whole technology to take.

Michael I. Jordan (54:19.920)

Be more limited, create, let humans rise up.

Lex Fridman (54:24.040)

Don't try to replace them.

Michael I. Jordan (54:25.740)

That's the AI mantra.

Lex Fridman (54:26.760)

Don't try to anticipate them.

Michael I. Jordan (54:28.660)

Don't try to predict them because you're, you're, you're not going to, you're not going

Lex Fridman (54:32.320)

to be able to do those things.

Michael I. Jordan (54:33.320)

You're going to make things worse.

Lex Fridman (54:34.320)

Okay.

Lex Fridman (54:35.320)

So right now, just give this a chance.

Michael I. Jordan (54:38.760)

Right now, the recommender systems are the creepy people in the shadow watching your

Michael I. Jordan (54:43.840)

every move.

Lex Fridman (54:45.500)

So they're looking at traces of you.

Michael I. Jordan (54:47.800)

They're not directly interacting with you, sort of the, your close friends and family,

Michael I. Jordan (54:53.000)

the way they know you is by having conversation, by actually having interactions back and forth.

Lex Fridman (54:57.120)

Do you think there's a place for recommender systems sort of to step, cause you, you just

Michael I. Jordan (55:02.360)

emphasize the value of human to human connection, but yeah, just give it a chance, AI human

Michael I. Jordan (55:06.740)

connection.

Michael I. Jordan (55:07.840)

Is there a role for an AI system to have conversations with you in terms of, to try to figure out

Lex Fridman (55:13.560)

what kind of music you like, not by just watching what you listening to, but actually having

Lex Fridman (55:17.360)

a conversation, natural language or otherwise.

Michael I. Jordan (55:19.560)

Yeah, no, I'm, I'm, so I'm not against it.

Michael I. Jordan (55:21.760)

I just wanted to push back against the, maybe you're saying you have options for Facebook.

Lex Fridman (55:25.120)

So there I think it's misplaced, but, but I think that distributing, yeah, no, so good

Lex Fridman (55:31.760)

for you.

Michael I. Jordan (55:33.520)

Go for it.

Lex Fridman (55:34.520)

That's a hard spot to be in.

Michael I. Jordan (55:35.520)

Yeah, no, good.

Michael I. Jordan (55:36.520)

Human interaction, like on our daily, the context around me in my own home is something

Michael I. Jordan (55:39.520)

that I don't want some big company to know about at all, but I would be more than happy

Lex Fridman (55:42.280)

to have technology help me with it.

Lex Fridman (55:44.200)

Which kind of technology?

Lex Fridman (55:45.200)

Well, you know, just, Alexa, Amazon, well, a good, Alexa's done right.

Lex Fridman (55:49.200)

And I think Alexa is a research platform right now more than anything else.

Lex Fridman (55:52.160)

But Alexa done right, you know, could do things like I, I leave the water running in my garden

Lex Fridman (55:56.480)

and I say, Hey, Alexa, the water's running in my garden.

Lex Fridman (55:59.200)

And even have Alexa figure out that that means when my wife comes home, that she should be

Michael I. Jordan (56:02.040)

told about that.

Lex Fridman (56:03.600)

That's a little bit of a reasoning.

Michael I. Jordan (56:04.600)

I would call that AI and by any kind of stretch, it's a little bit of reasoning and it actually

Lex Fridman (56:08.860)

kind of would make my life a little easier and better.

Lex Fridman (56:11.000)

And you know, I don't, I wouldn't call this a wow moment, but I kind of think that overall

Lex Fridman (56:14.600)

rises human happiness up to have that kind of thing.

Lex Fridman (56:18.320)

But not when you're lonely, Alexa, knowing loneliness.

Lex Fridman (56:20.840)

No, no, I don't want Alexa to be, feel intrusive.

Lex Fridman (56:25.600)

And I don't want just the designer of the system to kind of work all this out.

Lex Fridman (56:28.440)

I really want to have a lot of control and I want transparency and control.

Lex Fridman (56:32.440)

And if a company can stand up and give me that in the context of new technology, I think

Lex Fridman (56:36.800)

they're good.

Michael I. Jordan (56:37.800)

First of all, be way more successful than our current generation.

Lex Fridman (56:39.280)

And like I said, I was mentioning Microsoft, I really think they're, they're pivoting to

Michael I. Jordan (56:43.300)

kind of be the trusted old uncle, but you know, I think that they get that this is a

Michael I. Jordan (56:47.000)

way to go, that if you let people find technology, empowers them to have more control and have

Lex Fridman (56:51.600)

and have control, not just over privacy, but over this rich set of interactions, that that

Lex Fridman (56:56.720)

people are going to like that a lot more.

Lex Fridman (56:58.120)

And that's, that's the right business model going forward.

Lex Fridman (57:00.560)

What does control over privacy look like?

Lex Fridman (57:02.240)

Do you think you should be able to just view all the data that?

Lex Fridman (57:04.760)

No, it's much more than that.

Michael I. Jordan (57:05.920)

I mean, first of all, it should be an individual decision.

Lex Fridman (57:07.900)

Some people don't want privacy.

Michael I. Jordan (57:09.220)

They want their whole life out there.

Lex Fridman (57:10.720)

Other people's want it.

Michael I. Jordan (57:13.720)

Privacy is not a zero one.

Lex Fridman (57:16.020)

It's not a legal thing.

Michael I. Jordan (57:17.020)

It's not just about which data is available, which is not.

Michael I. Jordan (57:20.280)

I like to recall to people that, you know, a couple hundred years ago, everyone, there

Michael I. Jordan (57:24.880)

was not really big cities, everyone lived in on the countryside and villages and villages.

Lex Fridman (57:29.640)

Everybody knew everything about you.

Michael I. Jordan (57:30.640)

Very, you didn't have any privacy.

Lex Fridman (57:32.720)

Is that bad?

Lex Fridman (57:33.720)

Are we better off now?

Michael I. Jordan (57:34.720)

Well, you know, arguably no, because what did you get for that loss of certain kinds

Lex Fridman (57:39.040)

of privacy?

Lex Fridman (57:40.520)

Well, people help each other if they, because they know everything about you.

Michael I. Jordan (57:44.080)

They know something's bad's happening, they will help you with that.

Lex Fridman (57:46.400)

Right.

Lex Fridman (57:47.400)

And now you live in a big city, no one knows about that.

Lex Fridman (57:48.400)

You get no help.

Lex Fridman (57:50.840)

So it kind of depends the answer.

Lex Fridman (57:52.680)

I want certain people who I trust and there should be relationships.

Lex Fridman (57:56.320)

I should kind of manage all those, but who knows what about me?

Lex Fridman (57:59.000)

I should have some agency there.

Michael I. Jordan (58:00.800)

It shouldn't, I shouldn't be a drift in a sea of technology where I have no agency.

Lex Fridman (58:04.680)

I don't want to go reading things and checking boxes.

Lex Fridman (58:08.560)

So I don't know how to do that.

Lex Fridman (58:09.960)

And I'm not a privacy researcher per se.

Michael I. Jordan (58:11.480)

I just, I recognize the vast complexity of this.

Lex Fridman (58:14.360)

It's not just technology.

Michael I. Jordan (58:15.360)

It's not just legal scholars meeting technologists.

Lex Fridman (58:18.920)

There's gotta be kind of a whole layers around it.

Lex Fridman (58:20.900)

And so I, when I alluded to this emerging engineering field, this is a big part of it.

Michael I. Jordan (58:26.480)

When electrical engineering came, I'm not one around at the time, but you just didn't

Michael I. Jordan (58:31.320)

plug electricity into walls and all kinds of work.

Michael I. Jordan (58:34.120)

You don't have to have like underwriters laboratory that reassured you that that plug's not going

Michael I. Jordan (58:37.840)

to burn up your house and that that machine will do this and that and everything.

Lex Fridman (58:41.720)

There'll be whole people who can install things.

Michael I. Jordan (58:44.520)

There'll be people who can watch the installers.

Lex Fridman (58:46.360)

There'll be a whole layers, you know, an onion of these kinds of things.

Lex Fridman (58:49.960)

And for things as deep and interesting as privacy, which is as least as interesting

Michael I. Jordan (58:53.960)

as electricity, that's going to take decades to kind of work out, but it's going to require

Michael I. Jordan (58:58.120)

a lot of new structures that we don't have right now.

Lex Fridman (59:00.320)

So it's kind of hard to talk about it.

Lex Fridman (59:02.320)

And you're saying there's a lot of money to be made if you get it right.

Lex Fridman (59:04.840)

So something you should look at.

Michael I. Jordan (59:05.840)

A lot of money to be made in all these things that provide human services and people recognize

Lex Fridman (59:09.560)

them as useful parts of their lives.

Lex Fridman (59:12.360)

So yeah.

Lex Fridman (59:14.280)

So yeah, the dialogue sometimes goes from the exuberant technologists to the no technology

Michael I. Jordan (59:19.660)

is good, kind of.

Lex Fridman (59:20.800)

And that's, you know, in our public discourse, you know, and as far as you see too much of

Michael I. Jordan (59:24.480)

this kind of thing and the sober discussions in the middle, which are the challenge he

Lex Fridman (59:28.400)

wants to have or where we need to be having our conversations.

Lex Fridman (59:31.560)

And you know, there's just not actually, there's not many forum fora for those.

Lex Fridman (59:36.480)

You know, there's, that's, that's kind of what I would look for.

Michael I. Jordan (59:39.180)

Maybe I could go and I could read a comment section of something and it would actually

Lex Fridman (59:42.040)

be this kind of dialogue going back and forth.

Lex Fridman (59:44.520)

You don't see much of this, right?

Michael I. Jordan (59:45.800)

Which is why actually there's a resurgence of podcasts out of all, because people are

Michael I. Jordan (59:49.800)

really hungry for conversation, but there's technology is not helping much.

Lex Fridman (59:55.760)

So comment sections of anything, including YouTube is not hurting and not helping.